diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..6c9b3f2
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,9 @@
+visual_res/*
+frames/
+__pycache__
+*.pt
+*.pth
+*.mp4
+*.pyc
+surgery_gesture_r2plus1d.py
+surgery*
\ No newline at end of file
diff --git a/README.md b/README.md
index 8b43444..b8e3ee7 100644
--- a/README.md
+++ b/README.md
@@ -1 +1,46 @@
 # VideoVisual
+
+This is a PyTorch demo implemented several visualization methods for video classification networks. The target is to provide a toolkit (as [TorchRay](https://github.com/facebookresearch/TorchRay) to image) to interprete commonly utilized video classfication networks, such as I3D, R(2+1)D, TSM et al., which is also called *attribution* task, namely the problem of determining which part of the input video is responsible for the value computed by a neural network.
+
+The current version supports attribution methods and video classification models as following:
+
+#### Video classification models:
+* **Pretrained on Kinetics-400**: I3D, R(2+1)D, R3D, MC3, TSM;
+* **Pretrained on EPIC-Kitchens**: (noun & verb): TSM.
+
+#### Attribution methods:
+* **Backprop-based**: Gradients, Gradients x Inputs, Integrated Gradients;
+* **Activation-based**: GradCAM (does not support TSM now);
+* **Perturbation-based**: Extremal Perturbation and Spatiotemporal Perturbation (An extension vision of extremal perturbation on video inputs).
+
+## Requirements
+
+* Python 3.6.5 or greater
+* PyTorch 1.2.0 or greater
+* matplotlib==2.2.3
+* numpy==1.14.3
+* opencv_python==4.1.2.30
+* torchvision==0.4.0a0
+* torchray==1.0.0.2
+* tqdm==4.45.0
+* pandas==0.23.3
+* scikit_image==0.15.0
+* Pillow==7.1.2
+* scikit_learn==0.22.2.post1
+
+## Running the code
+
+### Examples
+
+#### Saptiotemporal Perturbation + I3D (pretrained on Kinetics-400)
+`$ python main.py --videos_dir /home/acb11711tx/lzq/VideoVisual/test_data/kinetics/sampled_frames --model i3d --pretrain_dataset kinetics --vis_method perturb --num_iter 2000 --perturb_area 0.1`
+
+#### Spatiotemporal Perturbation + TSM (pretrained on EPIC-Kitchens-noun)
+`$ python main.py --videos_dir /home/acb11711tx/lzq/VideoVisual/test_data/epic-kitchens-noun/sampled_frames --model tsm --pretrain_dataset epic-kitchens-noun --vis_method perturb --num_iter 2000 --perturb_area 0.05`
+
+#### Integrated Gradients + R(2+1)D (pretrained on Kinetics-400)
+`$ python main.py --videos_dir /home/acb11711tx/lzq/VideoVisual/test_data/kinetics/sampled_frames --model r2plus1d --pretrain_dataset kinetics --vis_method integrated_grad`
+
+## License
+
+TorchRay is CC-BY-NC licensed, as found in the [LICENSE](LICENSE) file.
diff --git a/datasets/universal_dataset.py b/datasets/universal_dataset.py
new file mode 100644
index 0000000..88dc2ba
--- /dev/null
+++ b/datasets/universal_dataset.py
@@ -0,0 +1,74 @@
+import torch
+from torch import nn
+from torchvision import transforms
+from torch.utils.data import Dataset, DataLoader
+
+import os
+import json
+import numpy as np 
+from tqdm import tqdm
+from PIL import Image
+
+import sys
+sys.path.append(".")
+sys.path.append("..")
+from utils.LongRangeSample import long_range_sample
+
+class UniversalDataset (Dataset):
+    def __init__ (self, data_dir, model_name, class_namelist, clip_length=16):
+        self.data_dir = data_dir
+        self.model_name = model_name
+        self.class_namelist = class_namelist
+        self.clip_length = clip_length
+        
+        self.video_names = sorted(os.listdir(data_dir))
+        assert len(self.video_names) > 0, f'Given directory contains no video.'
+
+        if model_name == 'i3d':
+            self.transform = transforms.Compose([
+                    transforms.Resize((344, 256)),
+                    transforms.CenterCrop((224, 224)),
+                    transforms.ToTensor(),
+                    transforms.Normalize([0.43216, 0.39467, 0.37645], [0.22803, 0.22145, 0.21699]),
+                ])
+        elif model_name in ['r2plus1d', 'r3d', 'mc3']:
+            self.transform = transforms.Compose([
+                    transforms.Resize((172, 128)),
+                    transforms.CenterCrop((112, 112)),
+                    transforms.ToTensor(),
+                    transforms.Normalize([0.43216, 0.39467, 0.37645], [0.22803, 0.22145, 0.21699]),
+                ])
+        elif model_name in ['tsm', 'tsn']:  # mean = [0.485, 0.456, 0.406], std = [0.229, 0.224, 0.225]
+            self.transform = transforms.Compose([
+                    transforms.Resize((344, 256)),
+                    transforms.CenterCrop((224, 224)),
+                    transforms.ToTensor(),
+                    transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
+            ])
+
+    def __len__ (self):
+        return len(self.video_names)
+
+    def __getitem__ (self, idx):
+        video_name = self.video_names[idx]
+        if '.mp4' not in video_name:
+            video_frames_dir = os.path.join(self.data_dir, video_name)
+            frame_names = sorted([f for f in os.listdir(video_frames_dir) if '.png' in f or '.jpg' in f])
+            num_frame = len(frame_names)
+            assert num_frame > self.clip_length, \
+                f"Number of frames should be larger than {self.clip_length}, given {num_frame}"
+
+            clip_fidxs = long_range_sample(num_frame, self.clip_length, 'first')
+            clip_fidxs_tensor = torch.tensor(clip_fidxs).long()
+
+            clip_frames = [Image.open(os.path.join(video_frames_dir, 
+                                f'{fidx+1:09d}.png')) for fidx in clip_fidxs]
+            clip_tensor = torch.stack([self.transform(frame) for frame in clip_frames], dim=1)
+
+            label_name = video_name.split('-')[0]
+            label_idx = self.class_namelist.index(label_name)
+            return clip_tensor, label_idx, video_name, clip_fidxs_tensor
+        else:
+            raise Exception('Cannot process MP4 file yet.')
+            
+        
diff --git a/main.py b/main.py
new file mode 100644
index 0000000..afa9682
--- /dev/null
+++ b/main.py
@@ -0,0 +1,186 @@
+import argparse
+import torch
+import torchvision
+from torch import nn
+from torch.utils.data import Dataset, DataLoader
+
+import os
+import time
+import copy
+from tqdm import tqdm
+import pickle
+import numpy as np
+import pandas as pd
+import csv
+import json
+
+import warnings
+warnings.filterwarnings("ignore")
+
+import sys
+sys.path.append(".")
+sys.path.append("..")
+
+from path_dict import PathDict
+path_dict = PathDict()
+proj_root = path_dict.proj_root
+
+from utils.ImageShow import *
+
+from visual_meth.integrated_grad import integrated_grad
+from visual_meth.gradients import gradients
+from visual_meth.perturbation import video_perturbation
+from visual_meth.grad_cam import grad_cam
+
+parser = argparse.ArgumentParser()
+parser.add_argument("--videos_dir", type=str, default='')
+parser.add_argument("--model", type=str, default='r2plus1d',
+                    choices=['r2plus1d', 'r3d', 'mc3', 'i3d', 'tsn', 'trn', 'tsm'])
+parser.add_argument("--pretrain_dataset", type=str, default='kinetics',
+                    choices=['', 'kinetics', 'epic-kitchens-verb', 'epic-kitchens-noun'])
+parser.add_argument("--vis_method", type=str, default='integrated_grad',
+                    choices=['grad', 'grad*input', 'integrated_grad', 'grad_cam', 'perturb'])
+parser.add_argument("--save_label", type=str, default='')
+parser.add_argument("--no_gpu", action='store_true')
+parser.add_argument("--num_iter", type=int, default=1000)
+parser.add_argument('--polarity', type=str, default='positive',
+                    choices=['positive', 'negative'])
+parser.add_argument('--perturb_area', type=float, default=0.1,
+                    choices=[0.01, 0.02, 0.05, 0.1, 0.15, 0.2])
+args = parser.parse_args()
+
+# assert args.num_gpu >= -1
+# if args.num_gpu == 0:
+#     num_devices = 0
+#     multi_gpu = False
+#     device = torch.device("cpu")
+# elif args.num_gpu == 1:
+#     num_devices = 1
+#     multi_gpu = False
+#     device = torch.device("cuda")
+# elif args.num_gpu == -1:
+#     num_devices = torch.cuda.device_count()
+#     multi_gpu = (num_devices > 1)
+#     device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+# else:
+#     num_devices = args.num_gpu
+#     assert torch.cuda.device_count() >= num_devices, \
+#         f'Assign {args.num_gpu} GPUs, but only detected only {torch.cuda.device_count()} GPUs. Exiting...'
+#     multi_gpu = True
+#     device = torch.device("cuda")
+
+if args.no_gpu:
+    device = torch.device("cpu")
+    num_devices = 0
+else:
+    device = torch.device("cuda")
+    num_devices = 1
+
+assert os.path.isdir(args.videos_dir), \
+    f'Given directory of data does not exist: {args.videos_dir}.'
+
+if args.pretrain_dataset == 'kinetics':
+    if args.model == 'i3d':
+        from model_def.i3d import I3D as model
+        model_ft = model(num_classes=400)
+        i3d_pt_dir = os.path.join(proj_root, 'model_param/kinetics400_rgb_i3d.pth')
+        model_ft.load_state_dict(torch.load(i3d_pt_dir))
+        clip_length = 16
+    elif args.model == 'tsm':
+        from model_def.tsm import tsm as model
+        model_ft = model(400, segment_count=8, pretrained=args.pretrain_dataset)
+        clip_length = 8
+    else:   # Load pretrained models from PyTorch directly
+        clip_length = 16
+        if args.model == 'r2plus1d':
+            from torchvision.models.video import r2plus1d_18 as model
+        elif args.model == 'mc3':
+            from torchvision.models.video import mc3_18 as model
+        elif args.model == 'r3d':
+            from torchvision.models.video import r3d_18 as model
+        else:
+            raise Exception(f'Given model of {args.model} has no pretrain on {args.pretrain_dataset}.')
+        model_ft = model(pretrained=True)
+
+    model_ft = model_ft.to(device)
+    model_ft.eval()
+    # if multi_gpu:
+    #     model_ft = nn.DataParallel(model_ft, device_ids=list(range(num_devices)))
+
+    kinetics400_classes = os.path.join(proj_root, 'test_data/kinetics/classes.json')
+    class_namelist = json.load(open(kinetics400_classes))
+
+elif 'epic-kitchens' in args.pretrain_dataset:
+    if 'noun' in args.pretrain_dataset:
+        epic_classes = os.path.join(proj_root, 'test_data/epic-kitchens-noun/EPIC_noun_classes.csv')
+    elif 'verb' in args.pretrain_dataset:
+        epic_classes = os.path.join(proj_root, 'test_data/epic-kitchens-verb/EPIC_verb_classes.csv')
+    else:
+        raise Exception(f'EPIC-Kitchens only supports two sub-tasks (noun & verb), given {args.pretrain_dataset}.')
+    class_namelist = [row['class_key'] for ridx, row in pd.read_csv(epic_classes).iterrows()]
+    class_num = len(class_namelist)
+
+    if args.model == 'tsm':
+        from model_def.tsm import tsm as model
+        model_ft = model(class_num, segment_count=8, pretrained=args.pretrain_dataset)
+        clip_length = 16
+    else:
+        raise Exception(f'{args.pretrain_dataset} has only pretrained TSM model. Given {args.model}.')
+    model_ft = model_ft.to(device)
+    model_ft.eval()
+
+from datasets.universal_dataset import UniversalDataset as dataset
+test_dataset = dataset(args.videos_dir, args.model, class_namelist, clip_length=clip_length)
+test_dataloader = DataLoader(test_dataset, batch_size=1, shuffle=False)
+print(f'Num of test samples:{len(test_dataset)}')
+
+for sample in tqdm(test_dataloader):
+    inp = sample[0].to(device)
+    label = sample[1].to(dtype=torch.long)
+
+    inp_np = voxel_tensor_to_np(inp[0].detach().cpu())   # 3 x num_f x 224 224
+
+    if args.vis_method == 'integrated_grad':
+        res = integrated_grad(inp, label, model_ft, device, steps=50, polarity=args.polarity)
+        heatmap_np = res[0].numpy()
+    elif args.vis_method == 'grad':
+        res = gradients(inp, label, model_ft, device, polarity=args.polarity)
+        heatmap_np = res[0].numpy()
+    elif args.vis_method == 'grad*input':
+        res = gradients(inp, label, model_ft, device, multiply_input=True, polarity=args.polarity)
+        heatmap_np = res[0].numpy()
+    elif args.vis_method == 'grad_cam':
+        if args.model in ['i3d']:
+            layer_name = ['mixed_5c']
+        elif args.model in ['r2plus1d', 'mc3', 'r3d']:   # Load pretrained models from PyTorch directly
+            layer_name = ['layer4']
+        # elif args.model in ['tsm', 'tsn']:
+        #     layer_name = ['model', 'base_model', 'layer4']
+        else:
+            raise Exception(f'Grad-CAM does not support {args.model} currently')
+        res = grad_cam(inp, label, model_ft, device, layer_name=layer_name, norm_vis=True)
+        heatmap_np = overlap_maps_on_voxel_np(inp_np, res[0,0].cpu().numpy(), norm_map=False)
+    elif args.vis_method == 'perturb':
+        sigma = 11 if inp.shape[-1] == 112 else 23
+        res = video_perturbation(
+                    model_ft, inp, label, areas=[args.perturb_area], sigma=sigma, 
+                    max_iter=args.num_iter, variant="preserve",
+                    num_devices=num_devices, print_iter=100, perturb_type="fade")[0]
+        heatmap_np = overlap_maps_on_voxel_np(inp_np, res[0,0].cpu().numpy(), norm_map=False)
+
+    sample_name = sample[2][0].split("/")[-1]
+    plot_save_name = f"{sample_name}.png"
+    if args.vis_method in ['grad', 'grad*input', 'integrated_grad']:
+        plot_save_name = plot_save_name.replace('.png', f'{args.polarity}.png')
+    plot_save_dir = os.path.join(proj_root, "visual_res", args.vis_method, args.model)
+    if args.save_label != '':
+        plot_save_dir = os.path.join(plot_save_dir, args.save_label)
+    os.makedirs(plot_save_dir, exist_ok=True)
+
+    show_txt = f"{sample_name}"
+    plot_voxel_np(inp_np, heatmap_np, title=show_txt, 
+                    save_path=os.path.join(plot_save_dir, plot_save_name) )
+
+
+    
+
diff --git a/model_def/i3d.py b/model_def/i3d.py
new file mode 100644
index 0000000..7d0e8d9
--- /dev/null
+++ b/model_def/i3d.py
@@ -0,0 +1,424 @@
+import math
+import os
+
+import numpy as np
+import torch
+from torch.nn import ReplicationPad3d
+
+
+def get_padding_shape(filter_shape, stride):
+    def _pad_top_bottom(filter_dim, stride_val):
+        pad_along = max(filter_dim - stride_val, 0)
+        pad_top = pad_along // 2
+        pad_bottom = pad_along - pad_top
+        return pad_top, pad_bottom
+
+    padding_shape = []
+    for filter_dim, stride_val in zip(filter_shape, stride):
+        pad_top, pad_bottom = _pad_top_bottom(filter_dim, stride_val)
+        padding_shape.append(pad_top)
+        padding_shape.append(pad_bottom)
+    depth_top = padding_shape.pop(0)
+    depth_bottom = padding_shape.pop(0)
+    padding_shape.append(depth_top)
+    padding_shape.append(depth_bottom)
+
+    return tuple(padding_shape)
+
+
+def simplify_padding(padding_shapes):
+    all_same = True
+    padding_init = padding_shapes[0]
+    for pad in padding_shapes[1:]:
+        if pad != padding_init:
+            all_same = False
+    return all_same, padding_init
+
+
+class Unit3Dpy(torch.nn.Module):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 kernel_size=(1, 1, 1),
+                 stride=(1, 1, 1),
+                 activation='relu',
+                 padding='SAME',
+                 use_bias=False,
+                 use_bn=True):
+        super(Unit3Dpy, self).__init__()
+
+        self.padding = padding
+        self.activation = activation
+        self.use_bn = use_bn
+        if padding == 'SAME':
+            padding_shape = get_padding_shape(kernel_size, stride)
+            simplify_pad, pad_size = simplify_padding(padding_shape)
+            self.simplify_pad = simplify_pad
+        elif padding == 'VALID':
+            padding_shape = 0
+        else:
+            raise ValueError(
+                'padding should be in [VALID|SAME] but got {}'.format(padding))
+
+        if padding == 'SAME':
+            if not simplify_pad:
+                self.pad = torch.nn.ConstantPad3d(padding_shape, 0)
+                self.conv3d = torch.nn.Conv3d(
+                    in_channels,
+                    out_channels,
+                    kernel_size,
+                    stride=stride,
+                    bias=use_bias)
+            else:
+                self.conv3d = torch.nn.Conv3d(
+                    in_channels,
+                    out_channels,
+                    kernel_size,
+                    stride=stride,
+                    padding=pad_size,
+                    bias=use_bias)
+        elif padding == 'VALID':
+            self.conv3d = torch.nn.Conv3d(
+                in_channels,
+                out_channels,
+                kernel_size,
+                padding=padding_shape,
+                stride=stride,
+                bias=use_bias)
+        else:
+            raise ValueError(
+                'padding should be in [VALID|SAME] but got {}'.format(padding))
+
+        if self.use_bn:
+            self.batch3d = torch.nn.BatchNorm3d(out_channels)
+
+        if activation == 'relu':
+            self.activation = torch.nn.functional.relu
+
+    def forward(self, inp):
+        if self.padding == 'SAME' and self.simplify_pad is False:
+            inp = self.pad(inp)
+        out = self.conv3d(inp)
+        if self.use_bn:
+            out = self.batch3d(out)
+        if self.activation is not None:
+            out = torch.nn.functional.relu(out)
+        return out
+
+
+class MaxPool3dTFPadding(torch.nn.Module):
+    def __init__(self, kernel_size, stride=None, padding='SAME'):
+        super(MaxPool3dTFPadding, self).__init__()
+        if padding == 'SAME':
+            padding_shape = get_padding_shape(kernel_size, stride)
+            self.padding_shape = padding_shape
+            self.pad = torch.nn.ConstantPad3d(padding_shape, 0)
+        self.pool = torch.nn.MaxPool3d(kernel_size, stride, ceil_mode=True)
+
+    def forward(self, inp):
+        inp = self.pad(inp)
+        out = self.pool(inp)
+        return out
+
+
+class Mixed(torch.nn.Module):
+    def __init__(self, in_channels, out_channels):
+        super(Mixed, self).__init__()
+        # Branch 0
+        self.branch_0 = Unit3Dpy(
+            in_channels, out_channels[0], kernel_size=(1, 1, 1))
+
+        # Branch 1
+        branch_1_conv1 = Unit3Dpy(
+            in_channels, out_channels[1], kernel_size=(1, 1, 1))
+        branch_1_conv2 = Unit3Dpy(
+            out_channels[1], out_channels[2], kernel_size=(3, 3, 3))
+        self.branch_1 = torch.nn.Sequential(branch_1_conv1, branch_1_conv2)
+
+        # Branch 2
+        branch_2_conv1 = Unit3Dpy(
+            in_channels, out_channels[3], kernel_size=(1, 1, 1))
+        branch_2_conv2 = Unit3Dpy(
+            out_channels[3], out_channels[4], kernel_size=(3, 3, 3))
+        self.branch_2 = torch.nn.Sequential(branch_2_conv1, branch_2_conv2)
+
+        # Branch3
+        branch_3_pool = MaxPool3dTFPadding(
+            kernel_size=(3, 3, 3), stride=(1, 1, 1), padding='SAME')
+        branch_3_conv2 = Unit3Dpy(
+            in_channels, out_channels[5], kernel_size=(1, 1, 1))
+        self.branch_3 = torch.nn.Sequential(branch_3_pool, branch_3_conv2)
+
+    def forward(self, inp):
+        out_0 = self.branch_0(inp)
+        out_1 = self.branch_1(inp)
+        out_2 = self.branch_2(inp)
+        out_3 = self.branch_3(inp)
+        out = torch.cat((out_0, out_1, out_2, out_3), 1)
+        return out
+
+
+class I3D(torch.nn.Module):
+    def __init__(self,
+                 num_classes,
+                 modality='rgb',
+                 dropout_prob=0,
+                 name='inception'):
+        super(I3D, self).__init__()
+
+        self.name = name
+        self.num_classes = num_classes
+        if modality == 'rgb':
+            in_channels = 3
+        elif modality == 'flow':
+            in_channels = 2
+        else:
+            raise ValueError(
+                '{} not among known modalities [rgb|flow]'.format(modality))
+        self.modality = modality
+
+        conv3d_1a_7x7 = Unit3Dpy(
+            out_channels=64,
+            in_channels=in_channels,
+            kernel_size=(7, 7, 7),
+            stride=(2, 2, 2),
+            padding='SAME')
+        # 1st conv-pool
+        self.conv3d_1a_7x7 = conv3d_1a_7x7
+        self.maxPool3d_2a_3x3 = MaxPool3dTFPadding(
+            kernel_size=(1, 3, 3), stride=(1, 2, 2), padding='SAME')
+        # conv conv
+        conv3d_2b_1x1 = Unit3Dpy(
+            out_channels=64,
+            in_channels=64,
+            kernel_size=(1, 1, 1),
+            padding='SAME')
+        self.conv3d_2b_1x1 = conv3d_2b_1x1
+        conv3d_2c_3x3 = Unit3Dpy(
+            out_channels=192,
+            in_channels=64,
+            kernel_size=(3, 3, 3),
+            padding='SAME')
+        self.conv3d_2c_3x3 = conv3d_2c_3x3
+        self.maxPool3d_3a_3x3 = MaxPool3dTFPadding(
+            kernel_size=(1, 3, 3), stride=(1, 2, 2), padding='SAME')
+
+        # Mixed_3b
+        self.mixed_3b = Mixed(192, [64, 96, 128, 16, 32, 32])
+        self.mixed_3c = Mixed(256, [128, 128, 192, 32, 96, 64])
+
+        self.maxPool3d_4a_3x3 = MaxPool3dTFPadding(
+            kernel_size=(3, 3, 3), stride=(2, 2, 2), padding='SAME')
+
+        # Mixed 4
+        self.mixed_4b = Mixed(480, [192, 96, 208, 16, 48, 64])
+        self.mixed_4c = Mixed(512, [160, 112, 224, 24, 64, 64])
+        self.mixed_4d = Mixed(512, [128, 128, 256, 24, 64, 64])
+        self.mixed_4e = Mixed(512, [112, 144, 288, 32, 64, 64])
+        self.mixed_4f = Mixed(528, [256, 160, 320, 32, 128, 128])
+
+        self.maxPool3d_5a_2x2 = MaxPool3dTFPadding(
+            kernel_size=(2, 2, 2), stride=(2, 2, 2), padding='SAME')
+
+        # Mixed 5
+        self.mixed_5b = Mixed(832, [256, 160, 320, 32, 128, 128])
+        self.mixed_5c = Mixed(832, [384, 192, 384, 48, 128, 128])
+
+        self.avg_pool = torch.nn.AvgPool3d((2, 7, 7), (1, 1, 1))
+        self.dropout = torch.nn.Dropout(dropout_prob)
+        self.conv3d_0c_1x1 = Unit3Dpy(
+            in_channels=1024,
+            out_channels=self.num_classes,
+            kernel_size=(1, 1, 1),
+            activation=None,
+            use_bias=True,
+            use_bn=False)
+        self.softmax = torch.nn.Softmax(1)
+
+    def forward(self, inp):
+        # Preprocessing
+        out = self.conv3d_1a_7x7(inp)
+        out = self.maxPool3d_2a_3x3(out)
+        out = self.conv3d_2b_1x1(out)
+        out = self.conv3d_2c_3x3(out)
+        out = self.maxPool3d_3a_3x3(out)
+        out = self.mixed_3b(out)
+        out = self.mixed_3c(out)
+        out = self.maxPool3d_4a_3x3(out)
+        out = self.mixed_4b(out)
+        out = self.mixed_4c(out)
+        out = self.mixed_4d(out)
+        out = self.mixed_4e(out)
+        out = self.mixed_4f(out)
+        out = self.maxPool3d_5a_2x2(out)
+        out = self.mixed_5b(out)
+        out = self.mixed_5c(out)
+        out = self.avg_pool(out)
+        out = self.dropout(out)
+        out = self.conv3d_0c_1x1(out)
+        out = out.squeeze(3)
+        out = out.squeeze(3)
+        out = out.mean(2)
+        # out_logits = out
+        # out = self.softmax(out_logits)
+        # return out, out_logits
+        return out
+
+    def load_tf_weights(self, sess):
+        state_dict = {}
+        if self.modality == 'rgb':
+            prefix = 'RGB/inception_i3d'
+        elif self.modality == 'flow':
+            prefix = 'Flow/inception_i3d'
+        load_conv3d(state_dict, 'conv3d_1a_7x7', sess,
+                    os.path.join(prefix, 'Conv3d_1a_7x7'))
+        load_conv3d(state_dict, 'conv3d_2b_1x1', sess,
+                    os.path.join(prefix, 'Conv3d_2b_1x1'))
+        load_conv3d(state_dict, 'conv3d_2c_3x3', sess,
+                    os.path.join(prefix, 'Conv3d_2c_3x3'))
+
+        load_mixed(state_dict, 'mixed_3b', sess,
+                   os.path.join(prefix, 'Mixed_3b'))
+        load_mixed(state_dict, 'mixed_3c', sess,
+                   os.path.join(prefix, 'Mixed_3c'))
+        load_mixed(state_dict, 'mixed_4b', sess,
+                   os.path.join(prefix, 'Mixed_4b'))
+        load_mixed(state_dict, 'mixed_4c', sess,
+                   os.path.join(prefix, 'Mixed_4c'))
+        load_mixed(state_dict, 'mixed_4d', sess,
+                   os.path.join(prefix, 'Mixed_4d'))
+        load_mixed(state_dict, 'mixed_4e', sess,
+                   os.path.join(prefix, 'Mixed_4e'))
+        # Here goest to 0.1 max error with tf
+        load_mixed(state_dict, 'mixed_4f', sess,
+                   os.path.join(prefix, 'Mixed_4f'))
+
+        load_mixed(
+            state_dict,
+            'mixed_5b',
+            sess,
+            os.path.join(prefix, 'Mixed_5b'),
+            fix_typo=True)
+        load_mixed(state_dict, 'mixed_5c', sess,
+                   os.path.join(prefix, 'Mixed_5c'))
+        load_conv3d(
+            state_dict,
+            'conv3d_0c_1x1',
+            sess,
+            os.path.join(prefix, 'Logits', 'Conv3d_0c_1x1'),
+            bias=True,
+            bn=False)
+        self.load_state_dict(state_dict)
+
+
+def get_conv_params(sess, name, bias=False):
+    # Get conv weights
+    conv_weights_tensor = sess.graph.get_tensor_by_name(
+        os.path.join(name, 'w:0'))
+    if bias:
+        conv_bias_tensor = sess.graph.get_tensor_by_name(
+            os.path.join(name, 'b:0'))
+        conv_bias = sess.run(conv_bias_tensor)
+    conv_weights = sess.run(conv_weights_tensor)
+    conv_shape = conv_weights.shape
+
+    kernel_shape = conv_shape[0:3]
+    in_channels = conv_shape[3]
+    out_channels = conv_shape[4]
+
+    conv_op = sess.graph.get_operation_by_name(
+        os.path.join(name, 'convolution'))
+    padding_name = conv_op.get_attr('padding')
+    padding = _get_padding(padding_name, kernel_shape)
+    all_strides = conv_op.get_attr('strides')
+    strides = all_strides[1:4]
+    conv_params = [
+        conv_weights, kernel_shape, in_channels, out_channels, strides, padding
+    ]
+    if bias:
+        conv_params.append(conv_bias)
+    return conv_params
+
+
+def get_bn_params(sess, name):
+    moving_mean_tensor = sess.graph.get_tensor_by_name(
+        os.path.join(name, 'moving_mean:0'))
+    moving_var_tensor = sess.graph.get_tensor_by_name(
+        os.path.join(name, 'moving_variance:0'))
+    beta_tensor = sess.graph.get_tensor_by_name(os.path.join(name, 'beta:0'))
+    moving_mean = sess.run(moving_mean_tensor)
+    moving_var = sess.run(moving_var_tensor)
+    beta = sess.run(beta_tensor)
+    return moving_mean, moving_var, beta
+
+
+def _get_padding(padding_name, conv_shape):
+    padding_name = padding_name.decode("utf-8")
+    if padding_name == "VALID":
+        return [0, 0]
+    elif padding_name == "SAME":
+        # return [math.ceil(int(conv_shape[0])/2), math.ceil(int(conv_shape[1])/2)]
+        return [
+            math.floor(int(conv_shape[0]) / 2),
+            math.floor(int(conv_shape[1]) / 2),
+            math.floor(int(conv_shape[2]) / 2)
+        ]
+    else:
+        raise ValueError('Invalid padding name ' + padding_name)
+
+
+def load_conv3d(state_dict, name_pt, sess, name_tf, bias=False, bn=True):
+    # Transfer convolution params
+    conv_name_tf = os.path.join(name_tf, 'conv_3d')
+    conv_params = get_conv_params(sess, conv_name_tf, bias=bias)
+    if bias:
+        conv_weights, kernel_shape, in_channels, out_channels, strides, padding, conv_bias = conv_params
+    else:
+        conv_weights, kernel_shape, in_channels, out_channels, strides, padding = conv_params
+
+    conv_weights_rs = np.transpose(
+        conv_weights, (4, 3, 0, 1,
+                       2))  # to pt format (out_c, in_c, depth, height, width)
+    state_dict[name_pt + '.conv3d.weight'] = torch.from_numpy(conv_weights_rs)
+    if bias:
+        state_dict[name_pt + '.conv3d.bias'] = torch.from_numpy(conv_bias)
+
+    # Transfer batch norm params
+    if bn:
+        conv_tf_name = os.path.join(name_tf, 'batch_norm')
+        moving_mean, moving_var, beta = get_bn_params(sess, conv_tf_name)
+
+        out_planes = conv_weights_rs.shape[0]
+        state_dict[name_pt + '.batch3d.weight'] = torch.ones(out_planes)
+        state_dict[name_pt +
+                   '.batch3d.bias'] = torch.from_numpy(beta.squeeze())
+        state_dict[name_pt
+                   + '.batch3d.running_mean'] = torch.from_numpy(moving_mean.squeeze())
+        state_dict[name_pt
+                   + '.batch3d.running_var'] = torch.from_numpy(moving_var.squeeze())
+
+
+def load_mixed(state_dict, name_pt, sess, name_tf, fix_typo=False):
+    # Branch 0
+    load_conv3d(state_dict, name_pt + '.branch_0', sess,
+                os.path.join(name_tf, 'Branch_0/Conv3d_0a_1x1'))
+
+    # Branch .1
+    load_conv3d(state_dict, name_pt + '.branch_1.0', sess,
+                os.path.join(name_tf, 'Branch_1/Conv3d_0a_1x1'))
+    load_conv3d(state_dict, name_pt + '.branch_1.1', sess,
+                os.path.join(name_tf, 'Branch_1/Conv3d_0b_3x3'))
+
+    # Branch 2
+    load_conv3d(state_dict, name_pt + '.branch_2.0', sess,
+                os.path.join(name_tf, 'Branch_2/Conv3d_0a_1x1'))
+    if fix_typo:
+        load_conv3d(state_dict, name_pt + '.branch_2.1', sess,
+                    os.path.join(name_tf, 'Branch_2/Conv3d_0a_3x3'))
+    else:
+        load_conv3d(state_dict, name_pt + '.branch_2.1', sess,
+                    os.path.join(name_tf, 'Branch_2/Conv3d_0b_3x3'))
+
+    # Branch 3
+    load_conv3d(state_dict, name_pt + '.branch_3.1', sess,
+                os.path.join(name_tf, 'Branch_3/Conv3d_0b_1x1'))
diff --git a/model_def/tsm.py b/model_def/tsm.py
new file mode 100644
index 0000000..6fd637e
--- /dev/null
+++ b/model_def/tsm.py
@@ -0,0 +1,31 @@
+import torch
+from torch import nn
+
+class tsm (nn.Module):
+    def __init__ (self, num_classes, segment_count, pretrained):
+        super(tsm, self).__init__()
+        self.pretrained = pretrained
+        self.repo = 'epic-kitchens/action-models'
+        if 'epic-kitchens' in self.pretrained:
+            all_classes_num = (125, 352)
+            self.model = torch.hub.load(self.repo, 'TSM', all_classes_num, segment_count, 'RGB',
+                                            base_model='resnet50', pretrained='epic-kitchens')
+        elif 'kinetics' in self.pretrained:
+            kinetics_classes_num = 400
+            self.model = torch.hub.load(self.repo, 'TSM', kinetics_classes_num, segment_count, 'RGB',
+                                            base_model='resnet50')
+            checkpoint_path = '/home/acb11711tx/lzq/ModelVisualization/model_param/TSM_kinetics_RGB_resnet50_shift8_blockres_avg_segment8_e100_dense.pth'
+            state_dict = torch.load(checkpoint_path)['state_dict']
+            state_dict = {k[7:]: v for k, v in state_dict.items()}
+            self.model.load_state_dict(state_dict)
+
+    def forward (self, inp):
+        if 'epic-kitchens' in self.pretrained:
+            feat = self.model.features(inp)
+            verb_logits, noun_logits = self.model.logits(feat)
+            if 'noun' in self.pretrained:
+                return noun_logits
+            elif 'verb' in self.pretrained:
+                return verb_logits
+        elif 'kinetics' in self.pretrained:
+            return self.model(inp)
\ No newline at end of file
diff --git a/path_dict.py b/path_dict.py
new file mode 100644
index 0000000..43f8c6b
--- /dev/null
+++ b/path_dict.py
@@ -0,0 +1,14 @@
+import socket
+import os
+
+class PathDict (object):
+    def __init__ (self):
+        hostname = socket.gethostname()
+        if 'abci' in hostname:
+            self.surgery_ds_root = '/groups1/gcb50205/lzq/dataset/JIGSAWS'
+            self.proj_root = os.path.dirname(os.path.abspath(__file__))
+            print(self.proj_root)
+        else:
+            # self.surgery_ds_root = '/home/shinkyo/dataset/JIGSAWS'
+            # self.proj_root = '/home/shinkyo/lzq/ModelVisualization'
+            self.proj_root = os.path.dirname(os.path.abspath(__file__))
diff --git a/requirements.txt b/requirements.txt
new file mode 100644
index 0000000..2545b3f
--- /dev/null
+++ b/requirements.txt
@@ -0,0 +1,12 @@
+matplotlib==2.2.3
+opencv_python==4.1.2.30
+torchvision==0.4.0a0
+torchray==1.0.0.2
+tqdm==4.45.0
+pandas==0.23.3
+scikit_image==0.15.0
+numpy==1.14.3
+torch==1.2.0
+Pillow==7.1.2
+scikit_learn==0.22.2.post1
+skimage==0.0
diff --git a/test_data/epic-kitchens-noun/EPIC_noun_classes.csv b/test_data/epic-kitchens-noun/EPIC_noun_classes.csv
new file mode 100644
index 0000000..9be7c63
--- /dev/null
+++ b/test_data/epic-kitchens-noun/EPIC_noun_classes.csv
@@ -0,0 +1,353 @@
+noun_id,class_key,nouns
+0,Nothing,['Nothing']
+1,pan,"['pan', 'pan:sauce', 'pan:frying', 'pan:cake', 'saucepan', 'saucepan:empty', 'saucepan:small', 'wok', 'wok:scour', 'pan:content']"
+2,pan:dust,"['pan:dust', 'dustpan']"
+3,tap,"['tap', 'tap:water', 'nozzle', 'nozzle:coffee']"
+4,plate,"['plate', 'plate:pasta', 'saucer', 'dish', 'plate:salad', 'plate:lasagna', 'food:plate', 'flatware', 'pate']"
+5,knife,"['knife', 'knife:mincing', 'knife:using', 'blade', 'chopper', 'knife:mezzaluna', 'chopper:moon', 'mezzaluna']"
+6,bowl,"['bowl', 'bowl:salad', 'bowl:glass', 'bowl:sugar', 'bowl:empty', 'bowl:cooker', 'bowl:washing', 'bowl:egg', 'bowl:processor']"
+7,spoon,"['spoon', 'spoon:wooden', 'spoon:using', 'teaspoon', 'risotto:spoon', 'spade', 'spade:food', 'spoon:coffee', 'spoon:tea', 'measure', 'stirrer:coffee', 'scoop', 'stirrer', 'tablespoon']"
+8,cupboard,"['cupboard', 'cabinet', 'locker', 'flap', 'door', 'door:cabinet', 'door:cupboard', 'closet']"
+9,drawer,"['drawer', 'drawer:dishwasher', 'drawer:freezer', 'drawer:refrigerator', 'draw']"
+10,fridge,"['fridge', 'refrigerator', 'door:refrigerator', 'door:fridge']"
+11,lid,"['lid', 'lid:pot', 'lid:tin', 'lid:saucepan', 'lid:tupperware', 'lid:box', 'lid:pan', 'lid:bottle', 'lid:teapot', 'lid:kettle', 'lid:yoghurt', 'lid:chopper', 'seal']"
+12,hand,"['hand', 'hand:using', 'hand:my']"
+13,onion,"['onion', 'onion:cutting', 'piece:onion', 'bit:onion']"
+14,onion:spring,['onion:spring']
+15,pot,"['pot', 'pot:coffee', 'pot:small', 'pot:white', 'pot:right', 'pot:left', 'pot:teapot', 'pot:second', 'teapot', 'part:teapot', 'pot:pressure', 'pot:pasta', 'pot:sauce', 'pot:third', 'pot:first']"
+16,glass,"['glass', 'glass:wine', 'glass:empty']"
+17,water,"['water', 'water:tap', 'water:empty', 'water:excess', 'water:hot', 'water:more', 'excess:water']"
+18,fork,"['fork', 'fork:soap']"
+19,board:chopping,"['board:chopping', 'board:cutting', 'board', 'board:serving', 'block:chopping', 'board:cut', 'block']"
+20,bag,"['bag', 'bag:plastic', 'bag:pasta', 'bag:mozzarella', 'bag:cereal', 'bag:grape', 'bag:flour', 'bag:cheese', 'bag:nut', 'bag:rice', 'bag:oregano', 'bag:salad', 'bag:bin', 'bag:coffee', 'bag:spaghetti', 'bag:food', 'bag:trash', 'bag:onion', 'bag:carrot', 'bag:bread', 'bag:grapes', 'bag:sandwich', 'bag:kitchen']"
+21,sponge,"['sponge', 'scrubber']"
+22,spatula,"['spatula', 'spatula:wooden', 'spate']"
+23,cup,"['cup', 'cup:coffee', 'cup:measuring', 'cup:small', 'mug', 'mug:rinse', 'mug:clean']"
+24,oil,"['oil', 'oil:olive', 'oil:chilli', 'oil:sesame']"
+25,bin,"['bin', 'bin:trash', 'bin:dust', 'bin:food', 'dustbin', 'bin:rubbish', 'bin:garbage', 'bin:recycling', 'basket:waste', 'wastebasket']"
+26,meat,"['meat', 'meat:minced', 'piece:meat', 'mix:meat', 'meat:mix', 'mincemeat', 'meat:mince']"
+27,potato,"['potato', 'potato:baby']"
+28,bottle,"['bottle', 'bottle:oil', 'bottle:milk', 'bottle:water', 'bottle:sauce', 'bottle:pesto', 'bottle:vinegar']"
+29,container,"['container', 'container:food', 'container:milk', 'container:sauce', 'container:rice', 'container:tofu', 'container:sugar', 'container:salt', 'container:tomato', 'container:pasta', 'container:cake', 'bin:bread', 'bucket']"
+30,tomato,"['tomato', 'tomato:dried', 'tomato:cherry', 'tomato:chopped', 'sun']"
+31,salt,"['salt', 'seasoning']"
+32,cloth,"['cloth', 'cloth:table', 'rag', 'cloth:dish', 'cloth:cleaning', 'cloth:washing', 'dishrag']"
+33,sink,"['sink', 'sink:kitchen', 'sink:water']"
+34,door:kitchen,['door:kitchen']
+35,pasta,"['pasta', 'spaghetti', 'piece:pasta', 'lasagne']"
+36,dish:soap,['dish:soap']
+37,food,"['food', 'food:scoop', 'residue:food', 'bit:food', 'grocery']"
+38,kettle,['kettle']
+39,box,"['box', 'box:lunch', 'box:cereal', 'box:tea', 'box:salt', 'tupperware', 'lunchbox', 'case', 'box:mushroom', 'box:coffee', 'box:cheese', 'box:milk', 'box:rice', 'box:carrot', 'box:fruit', 'pot:plastic', 'tub', 'box:plastic', 'box:eggs', 'box:fajita']"
+40,carrot,"['carrot', 'carrot:tail']"
+41,sauce,"['sauce', 'sauce:soy', 'sauce:fish', 'sauce:pesto', 'sauce:tomato', 'salsa', 'tabasco', 'sauce:soya', 'soy', 'passata', 'mirin']"
+42,colander,"['colander', 'strainer:pasta']"
+43,milk,"['milk', 'milk:soy', 'milk:coconut', 'milk:soya']"
+44,rice,"['rice', 'rice:scoop']"
+45,garlic,"['garlic', 'garlic:chop', 'garlic:peel', 'piece:garlic', 'clove', 'bit:garlic', 'bulb']"
+46,pepper,"['pepper', 'pepper:bell']"
+47,hob,"['hob', 'stove', 'gas', 'burners', 'hop', 'hot', 'burner']"
+48,dough,['dough']
+49,dishwasher,"['dishwasher', 'dishwater']"
+50,egg,"['egg', 'flour:egg']"
+51,cheese,"['cheese', 'slice:cheese', 'mixture:cheese', 'mozzarella', 'paneer', 'slice:mozzarella', 'parmesan']"
+52,bread,"['bread', 'bread:white', 'bread:toast', 'loaf', 'rusk', 'toast', 'bun:hamburger', 'slice:bread', 'ciabatta', 'bread:sandwich']"
+53,table,"['table', 'table:round']"
+54,salad,"['salad', 'salad:cucumber']"
+55,microwave,"['microwave', 'oven:microwave']"
+56,oven,"['oven', 'oven:adjust', 'cooker', 'door:oven']"
+57,cooker:slow,"['cooker:slow', 'cooker:rice']"
+58,coffee,"['coffee', 'coffee:scoop', 'pod:coffee', 'bean:coffee', 'ground:coffee', 'scoop:coffee', 'spill:coffee', 'grounds', 'ground']"
+59,filter,"['filter', 'filter:water', 'filter:coffee', 'filter:empty', 'filter:sink', 'paper:filter', 'catcher:sink', 'stud']"
+60,jar,"['jar', 'jar:sugar', 'jar:coffee', 'jar:filter', 'jar:salt', 'jar:honey']"
+61,rack:drying,"['rack:drying', 'rack', 'rack:dish', 'drainer']"
+62,chicken,"['chicken', 'thigh:chicken', 'breast:chicken', 'piece:chicken', 'breast']"
+63,tray,"['tray', 'tray:baking', 'pan:baking', 'tray:pizza', 'tin:baking', 'base:tin']"
+64,mixture,"['mixture', 'mix']"
+65,towel,"['towel', 'towel:tea']"
+66,towel:kitchen,"['towel:kitchen', 'towel:paper', 'roll:kitchen']"
+67,peach,"['peach', 'slice:peach']"
+68,skin,"['skin', 'skin:onion', 'skin:potato', 'peel', 'peel:onion', 'peeling', 'peeling:onion', 'peeling:carrot', 'rind', 'skin:squash', 'skin:carrot', 'skin:banana']"
+69,courgette,"['courgette', 'piece:courgette']"
+70,liquid:washing,"['liquid:washing', 'liquid:cleaning', 'detergent', 'degreaser', 'cleanser', 'spray', 'cleaner', 'washing:dish', 'product']"
+71,liquid,['liquid']
+72,leaf,"['leaf', 'leaf:lettuce', 'leaf:salad', 'leaf:spinach', 'leaf:bay']"
+73,lettuce,"['lettuce', 'rocket']"
+74,leaf:mint,"['leaf:mint', 'leaf:curry']"
+75,cutlery,"['cutlery', 'knife:fork', 'silverware', 'tableware']"
+76,scissors,"['scissors', 'scissor']"
+77,package,"['package', 'package:cheese', 'package:bread', 'package:cooky', 'package:popcorn', 'package:mozzarella', 'package:parsley', 'package:salad', 'package:pasta', 'packaging', 'packaging:salad', 'packaging:bread', 'packet', 'pack:cheese', 'pack', 'pack:tortilla', 'pack:tea', 'carton:egg', 'wrapper', 'plastic', 'package:sausage', 'package:cookies', 'pack:soup', 'package:bacon', 'pack:oat', 'carton', 'packet:beef', 'packet:tea', 'package:meat', 'sachet', 'packet:cookies', 'packet:cheese', 'carton:milk', 'sachet:flavour', 'packet:sausage', 'package:cream', 'package:salt', 'pack:salad', 'pack:meat', 'pack:pasta', 'package:fajita', 'package:wrap', 'package:spice', 'package:egg', 'cardboard', 'wrapping']"
+78,top,"['top', 'top:counter', 'top:stove', 'surface', 'surface:hob', 'surface:wipe', 'counter', 'counter:kitchen', 'side', 'top:kitchen', 'tile']"
+79,spice,['spice']
+80,tortilla,['tortilla']
+81,paper,"['paper', 'paper:baking', 'paper:cooking', 'letter']"
+82,machine:washing,"['machine:washing', 'machine', 'washer', 'machine:washer', 'door:machine']"
+83,olive,['olive']
+84,sausage,"['sausage', 'chorizo']"
+85,glove:oven,"['glove:oven', 'glove', 'mitt', 'mitt:oven', 'mitten']"
+86,peeler:potato,"['peeler:potato', 'peeler', 'skinner:carrot', 'skinner', 'peeler:carrot']"
+87,can,"['can', 'tin', 'tin:tomato']"
+88,mat,"['mat', 'mat:place', 'board:eating', 'placemat']"
+89,mat:sushi,['mat:sushi']
+90,vegetable,['vegetable']
+91,wrap:plastic,"['wrap:plastic', 'foil:transparent', 'film:cling', 'film']"
+92,wrap,['wrap']
+93,flour,"['flour', 'mixture:flour', 'powder:flour']"
+94,cucumber,['cucumber']
+95,curry,"['curry', 'curry:fish', 'masala:bean', 'paneer:saag', 'masala']"
+96,cereal,"['cereal', 'shreddies', 'granola', 'bowl:cereal']"
+97,napkin,['napkin']
+98,soap,['soap']
+99,squash,['squash']
+100,fish,"['fish', 'piece:fish', 'skin:fish', 'fragment:fish', 'skin:salmon']"
+101,chilli,"['chilli', 'chilli:chop', 'chilli:red', 'chilli:slice', 'pepper:chilli']"
+102,cover,"['cover', 'cover:plastic', 'cover:plate', 'cover:dish']"
+103,sugar,['sugar']
+104,aubergine,['aubergine']
+105,jug,"['jug', 'jug:measuring', 'pitcher', 'carafe', 'carafe:rinse']"
+106,heat,"['heat', 'temperature', 'temperature:gas', 'temperature:cooker', 'heat:cooker', 'temp']"
+107,leek,"['leek', 'piece:leek']"
+108,rubbish,"['rubbish', 'trash', 'garbage', 'waste']"
+109,ladle,['ladle']
+110,mushroom,['mushroom']
+111,stock,"['stock', 'cube:stock']"
+112,freezer,['freezer']
+113,light,['light']
+114,pizza,"['pizza', 'slice:pizza']"
+115,ball,['ball']
+116,yoghurt,['yoghurt']
+117,chopstick,"['chopstick', 'stick:chop']"
+118,grape,['grape']
+119,ginger,"['ginger', 'piece:ginger', 'galangal', 'root:galangal']"
+120,banana,['banana']
+121,oregano,['oregano']
+122,tuna,"['tuna', 'mixture:tuna']"
+123,kitchen,['kitchen']
+124,salmon,['salmon']
+125,basket,"['basket', 'hamper']"
+126,maker:coffee,"['maker:coffee', 'v60', 'cafetiere', 'cafetiere:rinse', 'maker', 'machine:coffee', 'pot:moka', 'pot:mocha', 'moka']"
+127,roll,['roll']
+128,brush,"['brush', 'brusher', 'brush:pastry', 'broom', 'wool', 'wool:steel']"
+129,lemon,"['lemon', 'slice:lemon']"
+130,clothes,"['clothes', 'laundry', 'washing:washing', 'washing']"
+131,grater,"['grater', 'grater:cheese']"
+132,strainer,"['strainer', 'strainer:mesh', 'sieve']"
+133,bacon,['bacon']
+134,avocado,['avocado']
+135,blueberry,['blueberry']
+136,pesto,['pesto']
+137,utensil,"['utensil', 'skimmer']"
+138,bean:green,"['bean:green', 'bean']"
+139,floor,['floor']
+140,lime,"['lime', 'piece:lime']"
+141,foil,"['foil', 'foil:kitchen', 'foil:aluminium']"
+142,grill,['grill']
+143,ingredient,"['ingredient', 'ingredient:pan']"
+144,scale,"['scale', 'system:weighing', 'scales', 'balance']"
+145,paste:garlic,"['paste:garlic', 'paste:miso', 'paste', 'paste:shrimp', 'paste:pepper', 'paste:nut']"
+146,processor:food,"['processor:food', 'processor', 'bit:processor']"
+147,nut:pine,"['nut:pine', 'nut', 'peanut', 'pine']"
+148,butter,['butter']
+149,butter:peanut,['butter:peanut']
+150,shelf,"['shelf', 'shelf:utensil']"
+151,timer,"['timer', 'timer:oven']"
+152,rinse,['rinse']
+153,tablecloth,['tablecloth']
+154,switch,['switch']
+155,powder:coconut,"['powder:coconut', 'powder:curry', 'powder:baking', 'powder', 'powder:garlic']"
+156,powder:washing,"['powder:washing', 'persil']"
+157,capsule,"['capsule', 'capsule:empty']"
+158,oat,['oat']
+159,tofu,"['tofu', 'curd:bean', 'curd']"
+160,lighter,['lighter']
+161,corn,"['corn', 'corn:sweet', 'cob:corn', 'cob', 'corncob', 'sweetcorn']"
+162,vinegar,"['vinegar', 'reduction:balsamic', 'vinegar:wine', 'vinegar:rice', 'reduction']"
+163,grinder,"['grinder', 'grinder:coffee']"
+164,cap,"['cap', 'cap:sink']"
+165,support,"['support', 'support:pan', 'trivet']"
+166,cream,"['cream', 'fraiche:creme', 'fraiche']"
+167,content,"['content', 'content:pan']"
+168,tongs,"['tongs', 'tong']"
+169,pie,"['pie', 'pastry']"
+170,fan:extractor,"['fan:extractor', 'fan', 'ventilator', 'hood']"
+171,raisin,['raisin']
+172,toaster,['toaster']
+173,broccoli,['broccoli']
+174,pin:rolling,"['pin:rolling', 'pin']"
+175,plug,"['plug', 'socket']"
+176,button,['button']
+177,tea,"['tea', 'bag:tea', 'teabag']"
+178,parsley,['parsley']
+179,flame,['flame']
+180,herb,['herb']
+181,base,"['base', 'base:pizza']"
+182,holder:filter,"['holder:filter', 'holder:pot', 'holder', 'holder:jug']"
+183,thyme,['thyme']
+184,honey,['honey']
+185,celery,['celery']
+186,kiwi,['kiwi']
+187,tissue,['tissue']
+188,time,"['time', 'time:cooking']"
+189,clip,"['clip', 'clip:bag']"
+190,noodle,['noodle']
+191,yeast,['yeast']
+192,hummus,"['hummus', 'humus']"
+193,coconut,['coconut']
+194,cabbage,['cabbage']
+195,spinach,['spinach']
+196,nutella,['nutella']
+197,fruit,['fruit']
+198,dressing:salad,"['dressing:salad', 'dressing']"
+199,omelette,"['omelette', 'tamagoyaki', 'frittata']"
+200,kale,['kale']
+201,paella,['paella']
+202,chip,['chip']
+203,opener:bottle,"['opener:bottle', 'opener:can', 'corkscrew', 'opener']"
+204,shirt,"['shirt', 't', 'shirt:tee']"
+205,chair,['chair']
+206,sandwich,['sandwich']
+207,burger:tuna,"['burger:tuna', 'patty', 'patty:tuna', 'burger', 'hamburger']"
+208,pancake,['pancake']
+209,leftover,"['leftover', 'food:leftover']"
+210,risotto,['risotto']
+211,pestle,['pestle']
+212,sock,['sock']
+213,pea,['pea']
+214,apron,['apron']
+215,juice,"['juice', 'juice:lemon']"
+216,wine,['wine']
+217,dust,['dust']
+218,desk,['desk']
+219,mesh,['mesh']
+220,oatmeal,['oatmeal']
+221,artichoke,['artichoke']
+222,remover:spot,['remover:spot']
+223,coriander,"['coriander', 'cilantro']"
+224,mocha,['mocha']
+225,quorn,['quorn']
+226,soup,['soup']
+227,turmeric,['turmeric']
+228,knob,"['knob', 'dial']"
+229,seed,"['seed', 'pit', 'seed:squash']"
+230,boxer,"['boxer', 'short:boxer']"
+231,paprika,['paprika']
+232,juicer:lime,['juicer:lime']
+233,guard:hand,"['guard:hand', 'guard']"
+234,apple,['apple']
+235,tahini,['tahini']
+236,finger,['finger']
+237,salami,"['salami', 'ham', 'prosciutto']"
+238,mayonnaise,['mayonnaise']
+239,biscuit,"['biscuit', 'cookie', 'cooky', 'cracker']"
+240,pear,['pear']
+241,mortar,"['mortar', 'bowl:mortar']"
+242,berry,['berry']
+243,beef,"['beef', 'steak']"
+244,squeezer:lime,['squeezer:lime']
+245,tail,['tail']
+246,stick:crab,['stick:crab']
+247,supplement,['supplement']
+248,phone,"['phone', 'mobile']"
+249,shell:egg,"['shell:egg', 'eggshell', 'shell']"
+250,pith,['pith']
+251,ring:onion,"['ring:onion', 'ring']"
+252,cherry,['cherry']
+253,cake,['cake']
+254,sprout,['sprout']
+255,almond,['almond']
+256,mint,['mint']
+257,flake:chilli,"['flake:chilli', 'flake']"
+258,cutter:pizza,"['cutter:pizza', 'cutter']"
+259,nesquik,['nesquik']
+260,blender,"['blender', 'mixer']"
+261,scrap,['scrap']
+262,backpack,['backpack']
+263,melon,['melon']
+264,breadcrumb,"['breadcrumb', 'crumb:bread']"
+265,sticker,['sticker']
+266,shrimp,['shrimp']
+267,smoothie,['smoothie']
+268,grass:lemon,['grass:lemon']
+269,ketchup,['ketchup']
+270,slicer,"['slicer', 'device:cutting']"
+271,stand,['stand']
+272,dumpling,['dumpling']
+273,watch,['watch']
+274,beer,['beer']
+275,power,['power']
+276,heater,['heater']
+277,basil,"['basil', 'plant:basil']"
+278,cinnamon,['cinnamon']
+279,crisp,['crisp']
+280,asparagus,['asparagus']
+281,drink,"['drink', 'soda']"
+282,fishcakes,"['fishcakes', 'cake:fish', 'fishcake']"
+283,mustard,['mustard']
+284,caper,['caper']
+285,whetstone,"['whetstone', 'sharpener:knife', 'sharpener', 'stone:sharpening', 'stone']"
+286,candle,['candle']
+287,control:remote,"['control:remote', 'control', 'remote']"
+288,instruction,['instruction']
+289,cork,"['cork', 'stopper']"
+290,tab,['tab']
+291,masher,['masher']
+292,part,['part']
+293,muffin,['muffin']
+294,shaker:pepper,['shaker:pepper']
+295,garni:bouquet,['garni:bouquet']
+296,popcorn,['popcorn']
+297,envelope,['envelope']
+298,chocolate,['chocolate']
+299,spot,['spot']
+300,window,['window']
+301,syrup,"['syrup', 'syrup:maple']"
+302,bar:cereal,"['bar:cereal', 'bar']"
+303,croissant,['croissant']
+304,coke,['coke']
+305,stereo,['stereo']
+306,alarm,['alarm']
+307,recipe,['recipe']
+308,handle,['handle']
+309,sleeve,['sleeve']
+310,cumin,"['cumin', 'jeera']"
+311,wire,['wire']
+312,label,['label']
+313,fire,['fire']
+314,presser,"['presser', 'squeezer', 'press', 'press:garlic', 'juicer']"
+315,air,['air']
+316,mouse,['mouse']
+317,boiler,['boiler']
+318,rest,"['rest', 'rest:spoon', 'holder:spoon']"
+319,tablet,['tablet']
+320,poster,['poster']
+321,trousers,['trousers']
+322,form,['form']
+323,rubber,"['rubber', 'band:rubber']"
+324,rug,['rug']
+325,sheets,"['sheets', 'sheet']"
+326,pepper:cayenne,['pepper:cayenne']
+327,waffle,['waffle']
+328,pineapple,['pineapple']
+329,turkey,['turkey']
+330,alcohol,['alcohol']
+331,rosemary,['rosemary']
+332,lead,['lead']
+333,book,"['book', 'geographic:national']"
+334,rim,['rim']
+335,gravy,['gravy']
+336,straw,['straw']
+337,hat,['hat']
+338,cd,['cd']
+339,slipper,['slipper']
+340,casserole,['casserole']
+341,ladder,['ladder']
+342,jambalaya,['jambalaya']
+343,wall,['wall']
+344,tube,['tube']
+345,lamp,['lamp']
+346,tarragon,['tarragon']
+347,heart,['heart']
+348,funnel,['funnel']
+349,whisk,['whisk']
+350,driver:screw,['driver:screw']
+351,trouser,['trouser']
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000001.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000001.png
new file mode 100644
index 0000000..727fb13
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000001.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000002.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000002.png
new file mode 100644
index 0000000..34f7d5b
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000002.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000003.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000003.png
new file mode 100644
index 0000000..db19c6b
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000003.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000004.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000004.png
new file mode 100644
index 0000000..7ed091c
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000004.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000005.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000005.png
new file mode 100644
index 0000000..45ffd10
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000005.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000006.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000006.png
new file mode 100644
index 0000000..ba9c0f2
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000006.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000007.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000007.png
new file mode 100644
index 0000000..97c69cf
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000007.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000008.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000008.png
new file mode 100644
index 0000000..feca898
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000008.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000009.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000009.png
new file mode 100644
index 0000000..8082397
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000009.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000010.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000010.png
new file mode 100644
index 0000000..36261d4
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000010.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000011.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000011.png
new file mode 100644
index 0000000..c224b71
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000011.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000012.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000012.png
new file mode 100644
index 0000000..6626fc6
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000012.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000013.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000013.png
new file mode 100644
index 0000000..1ba0e12
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000013.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000014.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000014.png
new file mode 100644
index 0000000..e4d6c98
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000014.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000015.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000015.png
new file mode 100644
index 0000000..2467d31
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000015.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000016.png b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000016.png
new file mode 100644
index 0000000..9c4e7db
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cloth-take-P12_06_18977/000000016.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000001.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000001.png
new file mode 100644
index 0000000..c2103b2
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000001.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000002.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000002.png
new file mode 100644
index 0000000..57c70fd
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000002.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000003.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000003.png
new file mode 100644
index 0000000..3229398
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000003.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000004.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000004.png
new file mode 100644
index 0000000..45bc937
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000004.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000005.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000005.png
new file mode 100644
index 0000000..0939e68
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000005.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000006.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000006.png
new file mode 100644
index 0000000..cea52ab
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000006.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000007.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000007.png
new file mode 100644
index 0000000..d710fb6
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000007.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000008.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000008.png
new file mode 100644
index 0000000..2ee937b
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000008.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000009.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000009.png
new file mode 100644
index 0000000..6b15a3f
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000009.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000010.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000010.png
new file mode 100644
index 0000000..82eaf01
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000010.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000011.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000011.png
new file mode 100644
index 0000000..d8c8aaf
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000011.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000012.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000012.png
new file mode 100644
index 0000000..9e179fb
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000012.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000013.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000013.png
new file mode 100644
index 0000000..fe9d762
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000013.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000014.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000014.png
new file mode 100644
index 0000000..cc90830
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000014.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000015.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000015.png
new file mode 100644
index 0000000..5aa576c
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000015.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000016.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000016.png
new file mode 100644
index 0000000..b67ce26
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P03_05_6777/000000016.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000001.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000001.png
new file mode 100644
index 0000000..0dc28ab
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000001.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000002.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000002.png
new file mode 100644
index 0000000..d4f5fe2
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000002.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000003.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000003.png
new file mode 100644
index 0000000..1803e6b
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000003.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000004.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000004.png
new file mode 100644
index 0000000..3a66539
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000004.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000005.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000005.png
new file mode 100644
index 0000000..bec3d22
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000005.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000006.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000006.png
new file mode 100644
index 0000000..f559795
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000006.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000007.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000007.png
new file mode 100644
index 0000000..67e8af4
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000007.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000008.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000008.png
new file mode 100644
index 0000000..8ce7274
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000008.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000009.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000009.png
new file mode 100644
index 0000000..543ab66
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000009.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000010.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000010.png
new file mode 100644
index 0000000..3965d1d
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000010.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000011.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000011.png
new file mode 100644
index 0000000..87e933a
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000011.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000012.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000012.png
new file mode 100644
index 0000000..2fd00a8
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000012.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000013.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000013.png
new file mode 100644
index 0000000..12c7ad9
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000013.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000014.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000014.png
new file mode 100644
index 0000000..fa148e6
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000014.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000015.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000015.png
new file mode 100644
index 0000000..b01bb80
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000015.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000016.png b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000016.png
new file mode 100644
index 0000000..59d9967
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/cupboard-open-P30_06_37623/000000016.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000001.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000001.png
new file mode 100644
index 0000000..d8bbe65
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000001.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000002.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000002.png
new file mode 100644
index 0000000..8593f70
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000002.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000003.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000003.png
new file mode 100644
index 0000000..e17dd07
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000003.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000004.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000004.png
new file mode 100644
index 0000000..c60eb82
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000004.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000005.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000005.png
new file mode 100644
index 0000000..7770efe
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000005.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000006.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000006.png
new file mode 100644
index 0000000..6442364
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000006.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000007.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000007.png
new file mode 100644
index 0000000..9f8ba81
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000007.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000008.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000008.png
new file mode 100644
index 0000000..705a2df
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000008.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000009.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000009.png
new file mode 100644
index 0000000..69ad206
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000009.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000010.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000010.png
new file mode 100644
index 0000000..dbe4e0c
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000010.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000011.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000011.png
new file mode 100644
index 0000000..98ac6b3
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000011.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000012.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000012.png
new file mode 100644
index 0000000..73937e5
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000012.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000013.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000013.png
new file mode 100644
index 0000000..c7f0115
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000013.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000014.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000014.png
new file mode 100644
index 0000000..30fcd88
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000014.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000015.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000015.png
new file mode 100644
index 0000000..1e3f2d2
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000015.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000016.png b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000016.png
new file mode 100644
index 0000000..b04bfc0
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/food-stir-P03_05_6751/000000016.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000001.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000001.png
new file mode 100644
index 0000000..0f9c443
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000001.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000002.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000002.png
new file mode 100644
index 0000000..07af028
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000002.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000003.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000003.png
new file mode 100644
index 0000000..f066079
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000003.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000004.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000004.png
new file mode 100644
index 0000000..ca1d926
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000004.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000005.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000005.png
new file mode 100644
index 0000000..34319fe
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000005.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000006.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000006.png
new file mode 100644
index 0000000..eea1e63
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000006.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000007.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000007.png
new file mode 100644
index 0000000..0ebbe45
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000007.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000008.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000008.png
new file mode 100644
index 0000000..fb24913
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000008.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000009.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000009.png
new file mode 100644
index 0000000..4de0737
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000009.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000010.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000010.png
new file mode 100644
index 0000000..9f47783
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000010.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000011.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000011.png
new file mode 100644
index 0000000..0023240
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000011.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000012.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000012.png
new file mode 100644
index 0000000..9447fb9
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000012.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000013.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000013.png
new file mode 100644
index 0000000..9fb7826
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000013.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000014.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000014.png
new file mode 100644
index 0000000..323981c
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000014.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000015.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000015.png
new file mode 100644
index 0000000..ca60d4c
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000015.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000016.png b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000016.png
new file mode 100644
index 0000000..8184cd9
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/lid-put-P17_03_20421/000000016.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000001.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000001.png
new file mode 100644
index 0000000..ef0a73c
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000001.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000002.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000002.png
new file mode 100644
index 0000000..deed2c2
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000002.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000003.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000003.png
new file mode 100644
index 0000000..9481098
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000003.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000004.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000004.png
new file mode 100644
index 0000000..877d7ec
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000004.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000005.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000005.png
new file mode 100644
index 0000000..911f141
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000005.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000006.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000006.png
new file mode 100644
index 0000000..8bedf7b
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000006.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000007.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000007.png
new file mode 100644
index 0000000..2981877
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000007.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000008.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000008.png
new file mode 100644
index 0000000..2481097
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000008.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000009.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000009.png
new file mode 100644
index 0000000..2bc3e0b
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000009.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000010.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000010.png
new file mode 100644
index 0000000..184c005
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000010.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000011.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000011.png
new file mode 100644
index 0000000..5a90f1d
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000011.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000012.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000012.png
new file mode 100644
index 0000000..c19113c
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000012.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000013.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000013.png
new file mode 100644
index 0000000..a36ccae
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000013.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000014.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000014.png
new file mode 100644
index 0000000..9915758
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000014.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000015.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000015.png
new file mode 100644
index 0000000..6fd411f
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000015.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000016.png b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000016.png
new file mode 100644
index 0000000..fab449a
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/package-tear-P17_03_20411/000000016.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000001.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000001.png
new file mode 100644
index 0000000..ed6dddf
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000001.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000002.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000002.png
new file mode 100644
index 0000000..f490d96
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000002.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000003.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000003.png
new file mode 100644
index 0000000..4d5aafa
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000003.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000004.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000004.png
new file mode 100644
index 0000000..259960f
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000004.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000005.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000005.png
new file mode 100644
index 0000000..819c52d
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000005.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000006.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000006.png
new file mode 100644
index 0000000..8aa03b2
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000006.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000007.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000007.png
new file mode 100644
index 0000000..fb3af8a
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000007.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000008.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000008.png
new file mode 100644
index 0000000..2187265
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000008.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000009.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000009.png
new file mode 100644
index 0000000..31296b6
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000009.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000010.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000010.png
new file mode 100644
index 0000000..9786f7f
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000010.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000011.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000011.png
new file mode 100644
index 0000000..021a94d
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000011.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000012.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000012.png
new file mode 100644
index 0000000..aabc3ef
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000012.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000013.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000013.png
new file mode 100644
index 0000000..5684f3d
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000013.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000014.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000014.png
new file mode 100644
index 0000000..4f3218e
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000014.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000015.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000015.png
new file mode 100644
index 0000000..7a84e56
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000015.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000016.png b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000016.png
new file mode 100644
index 0000000..c4c0f8a
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/pan-take-P26_21_32526/000000016.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000001.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000001.png
new file mode 100644
index 0000000..539908c
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000001.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000002.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000002.png
new file mode 100644
index 0000000..d13e584
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000002.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000003.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000003.png
new file mode 100644
index 0000000..477615e
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000003.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000004.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000004.png
new file mode 100644
index 0000000..97c404a
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000004.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000005.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000005.png
new file mode 100644
index 0000000..b7c47ac
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000005.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000006.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000006.png
new file mode 100644
index 0000000..61c0318
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000006.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000007.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000007.png
new file mode 100644
index 0000000..23bbeab
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000007.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000008.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000008.png
new file mode 100644
index 0000000..5e0ec32
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000008.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000009.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000009.png
new file mode 100644
index 0000000..22319b4
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000009.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000010.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000010.png
new file mode 100644
index 0000000..21b4e30
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000010.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000011.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000011.png
new file mode 100644
index 0000000..ded4bfb
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000011.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000012.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000012.png
new file mode 100644
index 0000000..1fb1de2
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000012.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000013.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000013.png
new file mode 100644
index 0000000..13ed40d
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000013.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000014.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000014.png
new file mode 100644
index 0000000..d57ef37
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000014.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000015.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000015.png
new file mode 100644
index 0000000..c9c47b0
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000015.png differ
diff --git a/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000016.png b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000016.png
new file mode 100644
index 0000000..e363543
Binary files /dev/null and b/test_data/epic-kitchens-noun/sampled_frames/spoon-put-P30_06_37608/000000016.png differ
diff --git a/test_data/epic-kitchens-verb/EPIC_verb_classes.csv b/test_data/epic-kitchens-verb/EPIC_verb_classes.csv
new file mode 100644
index 0000000..eda5d9b
--- /dev/null
+++ b/test_data/epic-kitchens-verb/EPIC_verb_classes.csv
@@ -0,0 +1,126 @@
+verb_id,class_key,verbs
+0,take,"['take', 'grab', 'pick', 'draw', 'get', 'grab-up', 'collect-from', 'take-up', 'grab-down', 'pull-down', 'fetch', 'pick-up']"
+1,put,"['put', 'pose', 'put-away', 'place-that', 'place-down', 'put', 'set-down', 'put-over', 'place-on', 'reture', 'put-on', 'place-away', 'swap', 'layer', 'position', 'put-down', 'put-back', 'place-back', 'return', 'pace', 'dose', 'put-up', 'leave', 'place', 'put-aside', 'lay', 'reposition', 'replace', 'put-onto']"
+2,open,"['open', 'unzip', 'open-up']"
+3,close,"['close', 'close-off', 'shut']"
+4,wash,"['wash', 'sponge', 'lather', 'wash-with', 'rinse', 'rinse-off', 'soap-up', 'wash-off', 'soap', 'rise', 'wash-up', 'clean-around', 'clean-off', 'clean-with', 'clean', 'clean-up', 'wipe-down', 'wipe-of', 'wipe', 'wipe-off']"
+5,cut,"['cut', 'chop', 'chop-off', 'cut-off', 'slice-into', 'slice-along', 'slice-off', 'stem', 'slice-up', 'cut-into', 'dice', 'half', 'halve', 'chop-up', 'snip', 'trim', 'slice']"
+6,mix,"['mix', 'beat', 'mix-around', 'stir-with', 'whisk', 'stir', 'blend', 'mix-in', 'stir-in']"
+7,pour,"['pour', 'pour-in', 'tip-in', 'pour-out', 'pour-on', 'pour-into', 'sieve']"
+8,throw,"['throw', 'throw-out', 'recycle', 'dispose-of', 'throw-over', 'throw-away', 'throw-in', 'bin', 'trow', 'toss']"
+9,move,"['move', 'transfer', 'move-around']"
+10,remove,"['remove', 'extract', 'take-off', 'remove-out', 'take-out', 'get-out', 'remove-inside', 'remove-from', 'pick-out']"
+11,dry,"['dry', 'dry-off', 'towel']"
+12,turn-on,"['turn-on', 'start', 'begin', 'ignite', 'switch-on', 'activate', 'water-on', 'play', 'start-to', 'restart', 'light']"
+13,turn,"['turn', 'rotate']"
+14,shake,"['shake', 'shake-off', 'shake-out']"
+15,turn-off,"['turn-off', 'switch-of', 'water-off', 'switch-off', 'switch-out', 'shut-off', 'turn-of']"
+16,peel,"['peel', 'skin-from', 'skin', 'peel-off', 'peel-back']"
+17,adjust,"['adjust', 'change', 'regulate']"
+18,empty,['empty']
+19,scoop,"['scoop', 'spoon-in', 'spoon', 'scoop-out', 'scoop-up']"
+20,check,"['check', 'ensure', 'test', 'look-in', 'watch', 'inspect', 'check-on']"
+21,squeeze,"['squeeze', 'squidge', 'squidge-into', 'squash', 'squish-into', 'wring-out', 'wring', 'squish', 'squeeze-into']"
+22,insert,"['insert', 'put-in', 'put-into', 'fit', 'place-in', 'put-inside']"
+23,press,"['press', 'push-down', 'collapse', 'compress', 'push', 'press-on']"
+24,fill,"['fill', 'fill-with', 'fill-up', 'stuff']"
+25,add,"['add', 'combine', 'add-to']"
+26,scrape,"['scrape', 'scrape-out', 'loose', 'scour', 'scrap', 'scrape-off']"
+27,sharpen,"['sharpen', 'thin']"
+28,wrap,"['wrap', 'top', 'lid', 'cover', 'seal', 'clamp', 'fasten', 'reseal', 'clip', 'wrap-up', 'tie']"
+29,roll,"['roll', 'roll-up']"
+30,sprinkle,"['sprinkle', 'sprinkle-on', 'drizzle', 'scatter', 'sprincle', 'crumble']"
+31,break,"['break', 'snap', 'crack', 'break-up']"
+32,flip,"['flip', 'overturn', 'turn-over']"
+33,hang,"['hang', 'drape', 'hang-up']"
+34,hold,['hold']
+35,sort,"['sort', 'rearrange', 'arrange', 'clear', 'tidy', 'line-up']"
+36,apply,"['apply', 'spread']"
+37,crush,"['crush', 'hit', 'grind', 'grate', 'mash', 'crush']"
+38,search,"['search', 'search-for', 'search-in', 'look-for', 'find', 'locate']"
+39,sample,"['sample', 'taste', 'smell', 'lick']"
+40,knead,['knead']
+41,set,"['set', 'set-up', 'set-out']"
+42,walk,"['walk', 'enter', 'enter-into', 'walk-into', 'walk-around', 'walk-to', 'walk-with']"
+43,divide,"['divide', 'split', 'detach', 'separate', 'distribute']"
+44,spray,"['spray', 'spay']"
+45,use,"['use', 'used-to', 'used']"
+46,fold,['fold']
+47,cook,"['cook', 'toast', 'fry-in', 'heat', 'fry', 'reduce', 'boil']"
+48,filter,"['filter', 'strain', 'drain', 'dump-out']"
+49,scrub,"['scrub', 'scrub-inside']"
+50,look,"['look', 'stare-at', 'see', 'read-on', 'read', 'look-at']"
+51,finish,"['finish', 'stop', 'end', 'do']"
+52,soak,"['soak', 'submerge', 'immerse']"
+53,brush,['brush']
+54,pull,"['pull', 'pull-out']"
+55,pat,"['pat', 'pat-into', 'dab', 'tap', 'tap-on', 'poke', 'pat-down']"
+56,form,"['form', 'shape', 'forge', 'make', 'shape-into']"
+57,measure,['measure']
+58,drink,"['drink', 'drink-from']"
+59,choose,"['choose', 'select']"
+60,serve,"['serve', 'plate-on', 'plate', 'dish', 'plate-up']"
+61,drop,"['drop', 'drop-on']"
+62,wear,['wear']
+63,rip,"['rip', 'tear', 'tear-off', 'tear-down']"
+64,tip,"['tip', 'tip-over', 'tip-out']"
+65,turn-down,['turn-down']
+66,gather,"['gather', 'collect']"
+67,eat,"['eat', 'chew-on', 'bite']"
+68,stack,"['stack', 'stack-up']"
+69,store,['store']
+70,switch,['switch']
+71,increase,"['increase', 'switch-up', 'turn-up']"
+72,carry,"['carry', 'have', 'bring']"
+73,lift,"['lift', 'raise', 'tilt', 'lift-up']"
+74,twist,"['twist', 'screw', 'screw-in', 'screw-on', 'tighten']"
+75,sweep,['sweep']
+76,rub,"['rub', 'rub-off']"
+77,unwrap,"['unwrap', 'unpack', 'unseal']"
+78,stab,['stab']
+79,attach,"['attach', 'connect', 'plug', 'assemble', 'plug-into', 'plug-in']"
+80,stretch,"['stretch', 'unfold']"
+81,lower,['lower']
+82,prepare,['prepare']
+83,unscrew,"['unscrew', 'untwist']"
+84,season,"['season', 'sweeten', 'pepper', 'salt']"
+85,video,['video']
+86,tap-off,['tap-off']
+87,set-off,['set-off']
+88,squirt,['squirt']
+89,load,['load']
+90,unroll,"['unroll', 'roll-out']"
+91,water,"['water', 'wet']"
+92,do,['do']
+93,flatten,"['flatten', 'flatten-with']"
+94,uncover,"['uncover', 'lid-off']"
+95,slide,"['slide', 'slide-out']"
+96,unplug,['unplug']
+97,level,['level']
+98,tear-out,['tear-out']
+99,feel,"['feel', 'touch']"
+100,fix,['fix']
+101,spill,['spill']
+102,pack,['pack']
+103,bake,['bake']
+104,blow,"['blow', 'blow-out']"
+105,sit-on,['sit-on']
+106,count,['count']
+107,dip,['dip']
+108,cool,['cool']
+109,flush,['flush']
+110,knife,['knife']
+111,fork,['fork']
+112,swirl,['swirl']
+113,stick,['stick']
+114,pet-down,['pet-down']
+115,realize,['realize']
+116,weigh,['weigh']
+117,defoliate,['defoliate']
+118,deseed,['deseed']
+119,tessellate,['tessellate']
+120,unfreeze,['unfreeze']
+121,decide-if,['decide-if']
+122,let-out,['let-out']
+123,save,['save']
+124,reverse,['reverse']
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000001.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000001.png
new file mode 100644
index 0000000..c2103b2
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000001.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000002.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000002.png
new file mode 100644
index 0000000..57c70fd
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000002.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000003.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000003.png
new file mode 100644
index 0000000..3229398
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000003.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000004.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000004.png
new file mode 100644
index 0000000..45bc937
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000004.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000005.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000005.png
new file mode 100644
index 0000000..0939e68
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000005.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000006.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000006.png
new file mode 100644
index 0000000..cea52ab
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000006.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000007.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000007.png
new file mode 100644
index 0000000..d710fb6
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000007.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000008.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000008.png
new file mode 100644
index 0000000..2ee937b
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000008.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000009.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000009.png
new file mode 100644
index 0000000..6b15a3f
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000009.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000010.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000010.png
new file mode 100644
index 0000000..82eaf01
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000010.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000011.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000011.png
new file mode 100644
index 0000000..d8c8aaf
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000011.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000012.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000012.png
new file mode 100644
index 0000000..9e179fb
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000012.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000013.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000013.png
new file mode 100644
index 0000000..fe9d762
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000013.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000014.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000014.png
new file mode 100644
index 0000000..cc90830
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000014.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000015.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000015.png
new file mode 100644
index 0000000..5aa576c
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000015.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000016.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000016.png
new file mode 100644
index 0000000..b67ce26
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P03_05_6777/000000016.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000001.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000001.png
new file mode 100644
index 0000000..0dc28ab
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000001.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000002.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000002.png
new file mode 100644
index 0000000..d4f5fe2
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000002.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000003.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000003.png
new file mode 100644
index 0000000..1803e6b
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000003.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000004.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000004.png
new file mode 100644
index 0000000..3a66539
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000004.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000005.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000005.png
new file mode 100644
index 0000000..bec3d22
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000005.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000006.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000006.png
new file mode 100644
index 0000000..f559795
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000006.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000007.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000007.png
new file mode 100644
index 0000000..67e8af4
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000007.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000008.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000008.png
new file mode 100644
index 0000000..8ce7274
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000008.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000009.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000009.png
new file mode 100644
index 0000000..543ab66
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000009.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000010.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000010.png
new file mode 100644
index 0000000..3965d1d
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000010.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000011.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000011.png
new file mode 100644
index 0000000..87e933a
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000011.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000012.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000012.png
new file mode 100644
index 0000000..2fd00a8
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000012.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000013.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000013.png
new file mode 100644
index 0000000..12c7ad9
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000013.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000014.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000014.png
new file mode 100644
index 0000000..fa148e6
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000014.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000015.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000015.png
new file mode 100644
index 0000000..b01bb80
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000015.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000016.png b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000016.png
new file mode 100644
index 0000000..59d9967
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/open-cupboard-P30_06_37623/000000016.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000001.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000001.png
new file mode 100644
index 0000000..0f9c443
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000001.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000002.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000002.png
new file mode 100644
index 0000000..07af028
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000002.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000003.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000003.png
new file mode 100644
index 0000000..f066079
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000003.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000004.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000004.png
new file mode 100644
index 0000000..ca1d926
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000004.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000005.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000005.png
new file mode 100644
index 0000000..34319fe
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000005.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000006.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000006.png
new file mode 100644
index 0000000..eea1e63
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000006.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000007.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000007.png
new file mode 100644
index 0000000..0ebbe45
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000007.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000008.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000008.png
new file mode 100644
index 0000000..fb24913
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000008.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000009.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000009.png
new file mode 100644
index 0000000..4de0737
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000009.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000010.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000010.png
new file mode 100644
index 0000000..9f47783
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000010.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000011.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000011.png
new file mode 100644
index 0000000..0023240
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000011.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000012.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000012.png
new file mode 100644
index 0000000..9447fb9
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000012.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000013.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000013.png
new file mode 100644
index 0000000..9fb7826
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000013.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000014.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000014.png
new file mode 100644
index 0000000..323981c
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000014.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000015.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000015.png
new file mode 100644
index 0000000..ca60d4c
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000015.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000016.png b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000016.png
new file mode 100644
index 0000000..8184cd9
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-lid-P17_03_20421/000000016.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000001.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000001.png
new file mode 100644
index 0000000..539908c
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000001.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000002.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000002.png
new file mode 100644
index 0000000..d13e584
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000002.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000003.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000003.png
new file mode 100644
index 0000000..477615e
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000003.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000004.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000004.png
new file mode 100644
index 0000000..97c404a
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000004.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000005.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000005.png
new file mode 100644
index 0000000..b7c47ac
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000005.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000006.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000006.png
new file mode 100644
index 0000000..61c0318
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000006.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000007.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000007.png
new file mode 100644
index 0000000..23bbeab
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000007.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000008.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000008.png
new file mode 100644
index 0000000..5e0ec32
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000008.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000009.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000009.png
new file mode 100644
index 0000000..22319b4
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000009.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000010.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000010.png
new file mode 100644
index 0000000..21b4e30
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000010.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000011.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000011.png
new file mode 100644
index 0000000..ded4bfb
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000011.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000012.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000012.png
new file mode 100644
index 0000000..1fb1de2
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000012.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000013.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000013.png
new file mode 100644
index 0000000..13ed40d
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000013.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000014.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000014.png
new file mode 100644
index 0000000..d57ef37
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000014.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000015.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000015.png
new file mode 100644
index 0000000..c9c47b0
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000015.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000016.png b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000016.png
new file mode 100644
index 0000000..e363543
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/put-spoon-P30_06_37608/000000016.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000001.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000001.png
new file mode 100644
index 0000000..d8bbe65
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000001.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000002.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000002.png
new file mode 100644
index 0000000..8593f70
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000002.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000003.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000003.png
new file mode 100644
index 0000000..e17dd07
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000003.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000004.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000004.png
new file mode 100644
index 0000000..c60eb82
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000004.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000005.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000005.png
new file mode 100644
index 0000000..7770efe
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000005.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000006.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000006.png
new file mode 100644
index 0000000..6442364
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000006.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000007.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000007.png
new file mode 100644
index 0000000..9f8ba81
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000007.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000008.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000008.png
new file mode 100644
index 0000000..705a2df
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000008.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000009.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000009.png
new file mode 100644
index 0000000..69ad206
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000009.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000010.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000010.png
new file mode 100644
index 0000000..dbe4e0c
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000010.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000011.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000011.png
new file mode 100644
index 0000000..98ac6b3
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000011.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000012.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000012.png
new file mode 100644
index 0000000..73937e5
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000012.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000013.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000013.png
new file mode 100644
index 0000000..c7f0115
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000013.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000014.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000014.png
new file mode 100644
index 0000000..30fcd88
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000014.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000015.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000015.png
new file mode 100644
index 0000000..1e3f2d2
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000015.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000016.png b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000016.png
new file mode 100644
index 0000000..b04bfc0
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/stir-food-P03_05_6751/000000016.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000001.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000001.png
new file mode 100644
index 0000000..727fb13
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000001.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000002.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000002.png
new file mode 100644
index 0000000..34f7d5b
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000002.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000003.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000003.png
new file mode 100644
index 0000000..db19c6b
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000003.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000004.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000004.png
new file mode 100644
index 0000000..7ed091c
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000004.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000005.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000005.png
new file mode 100644
index 0000000..45ffd10
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000005.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000006.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000006.png
new file mode 100644
index 0000000..ba9c0f2
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000006.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000007.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000007.png
new file mode 100644
index 0000000..97c69cf
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000007.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000008.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000008.png
new file mode 100644
index 0000000..feca898
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000008.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000009.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000009.png
new file mode 100644
index 0000000..8082397
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000009.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000010.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000010.png
new file mode 100644
index 0000000..36261d4
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000010.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000011.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000011.png
new file mode 100644
index 0000000..c224b71
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000011.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000012.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000012.png
new file mode 100644
index 0000000..6626fc6
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000012.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000013.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000013.png
new file mode 100644
index 0000000..1ba0e12
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000013.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000014.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000014.png
new file mode 100644
index 0000000..e4d6c98
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000014.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000015.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000015.png
new file mode 100644
index 0000000..2467d31
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000015.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000016.png b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000016.png
new file mode 100644
index 0000000..9c4e7db
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-cloth-P12_06_18977/000000016.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000001.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000001.png
new file mode 100644
index 0000000..ed6dddf
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000001.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000002.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000002.png
new file mode 100644
index 0000000..f490d96
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000002.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000003.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000003.png
new file mode 100644
index 0000000..4d5aafa
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000003.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000004.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000004.png
new file mode 100644
index 0000000..259960f
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000004.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000005.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000005.png
new file mode 100644
index 0000000..819c52d
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000005.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000006.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000006.png
new file mode 100644
index 0000000..8aa03b2
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000006.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000007.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000007.png
new file mode 100644
index 0000000..fb3af8a
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000007.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000008.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000008.png
new file mode 100644
index 0000000..2187265
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000008.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000009.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000009.png
new file mode 100644
index 0000000..31296b6
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000009.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000010.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000010.png
new file mode 100644
index 0000000..9786f7f
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000010.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000011.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000011.png
new file mode 100644
index 0000000..021a94d
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000011.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000012.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000012.png
new file mode 100644
index 0000000..aabc3ef
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000012.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000013.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000013.png
new file mode 100644
index 0000000..5684f3d
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000013.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000014.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000014.png
new file mode 100644
index 0000000..4f3218e
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000014.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000015.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000015.png
new file mode 100644
index 0000000..7a84e56
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000015.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000016.png b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000016.png
new file mode 100644
index 0000000..c4c0f8a
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/take-pan-P26_21_32526/000000016.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000001.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000001.png
new file mode 100644
index 0000000..ef0a73c
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000001.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000002.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000002.png
new file mode 100644
index 0000000..deed2c2
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000002.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000003.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000003.png
new file mode 100644
index 0000000..9481098
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000003.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000004.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000004.png
new file mode 100644
index 0000000..877d7ec
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000004.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000005.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000005.png
new file mode 100644
index 0000000..911f141
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000005.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000006.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000006.png
new file mode 100644
index 0000000..8bedf7b
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000006.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000007.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000007.png
new file mode 100644
index 0000000..2981877
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000007.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000008.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000008.png
new file mode 100644
index 0000000..2481097
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000008.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000009.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000009.png
new file mode 100644
index 0000000..2bc3e0b
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000009.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000010.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000010.png
new file mode 100644
index 0000000..184c005
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000010.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000011.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000011.png
new file mode 100644
index 0000000..5a90f1d
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000011.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000012.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000012.png
new file mode 100644
index 0000000..c19113c
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000012.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000013.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000013.png
new file mode 100644
index 0000000..a36ccae
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000013.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000014.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000014.png
new file mode 100644
index 0000000..9915758
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000014.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000015.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000015.png
new file mode 100644
index 0000000..6fd411f
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000015.png differ
diff --git a/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000016.png b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000016.png
new file mode 100644
index 0000000..fab449a
Binary files /dev/null and b/test_data/epic-kitchens-verb/sampled_frames/tear-package-P17_03_20411/000000016.png differ
diff --git a/test_data/frame_process.py b/test_data/frame_process.py
new file mode 100644
index 0000000..a097549
--- /dev/null
+++ b/test_data/frame_process.py
@@ -0,0 +1,50 @@
+import os
+from PIL import Image
+import cv2
+
+# basedir = '/home/acb11711tx/lzq/ModelVisualization/test_data/epic-kitchens-noun/frames'
+
+# # for video_name in os.listdir(basedir):
+# #     video_dir = os.path.join(basedir, video_name)
+# #     frame_names = sorted(os.listdir(video_dir))
+# #     for frame_index, frame_name in enumerate(frame_names):
+# #         frame_index = int(frame_name[6:-4])
+# #         os.system(f'mv {video_dir}/{frame_name} {video_dir}/{frame_index+1:09d}.jpg')
+# #     print(f'{video_name} is finished.')
+
+# for video_name in os.listdir(basedir):
+#     video_dir = os.path.join(basedir, video_name)
+#     frame_names = sorted(os.listdir(video_dir))
+#     num_frame = len(frame_names)
+#     for frame_index in range(num_frame):
+#         img = Image.open(f'{video_dir}/{frame_index+1:09d}.jpg')
+#         img.save(f'{video_dir}/{frame_index+1:09d}.png')
+#         os.system(f'rm {video_dir}/{frame_index+1:09d}.jpg')
+#     print(f'{video_name} is finished.')
+
+# import os
+import sys
+sys.path.append(".")
+sys.path.append("..")
+from utils.LongRangeSample import long_range_sample
+
+basedir = '/home/acb11711tx/lzq/VideoVisual/test_data'
+dataset_names = ['epic-kitchens-noun', 'epic-kitchens-verb', 'kinetics']
+for dataset in dataset_names:
+    ds_path = os.path.join(basedir, dataset, 'frames')
+    video_names = sorted(os.listdir(ds_path))
+    for video_name in video_names:
+        video_dir = os.path.join(ds_path, video_name)
+        frame_names = sorted(os.listdir(video_dir))
+        num_frame = len(frame_names)
+        sampled_fidxs = long_range_sample(num_frame, 16, 'first')
+
+        new_video_dir = video_dir.replace('frames', 'sampled_frames')
+        os.makedirs(new_video_dir, exist_ok=True)
+        for idx, fidx in enumerate(sampled_fidxs):
+            img = Image.open(f'{video_dir}/{fidx+1:09d}.png')
+            img = img.resize((344, 256))
+            img.save(f'{new_video_dir}/{idx+1:09d}.png')
+        print(f'{video_dir} finished!')
+
+    
\ No newline at end of file
diff --git a/test_data/kinetics/classes.json b/test_data/kinetics/classes.json
new file mode 100644
index 0000000..47e6bcf
--- /dev/null
+++ b/test_data/kinetics/classes.json
@@ -0,0 +1,402 @@
+[
+  "abseiling",
+  "air drumming",
+  "answering questions",
+  "applauding",
+  "applying cream",
+  "archery",
+  "arm wrestling",
+  "arranging flowers",
+  "assembling computer",
+  "auctioning",
+  "baby waking up",
+  "baking cookies",
+  "balloon blowing",
+  "bandaging",
+  "barbequing",
+  "bartending",
+  "beatboxing",
+  "bee keeping",
+  "belly dancing",
+  "bench pressing",
+  "bending back",
+  "bending metal",
+  "biking through snow",
+  "blasting sand",
+  "blowing glass",
+  "blowing leaves",
+  "blowing nose",
+  "blowing out candles",
+  "bobsledding",
+  "bookbinding",
+  "bouncing on trampoline",
+  "bowling",
+  "braiding hair",
+  "breading or breadcrumbing",
+  "breakdancing",
+  "brush painting",
+  "brushing hair",
+  "brushing teeth",
+  "building cabinet",
+  "building shed",
+  "bungee jumping",
+  "busking",
+  "canoeing or kayaking",
+  "capoeira",
+  "carrying baby",
+  "cartwheeling",
+  "carving pumpkin",
+  "catching fish",
+  "catching or throwing baseball",
+  "catching or throwing frisbee",
+  "catching or throwing softball",
+  "celebrating",
+  "changing oil",
+  "changing wheel",
+  "checking tires",
+  "cheerleading",
+  "chopping wood",
+  "clapping",
+  "clay pottery making",
+  "clean and jerk",
+  "cleaning floor",
+  "cleaning gutters",
+  "cleaning pool",
+  "cleaning shoes",
+  "cleaning toilet",
+  "cleaning windows",
+  "climbing a rope",
+  "climbing ladder",
+  "climbing tree",
+  "contact juggling",
+  "cooking chicken",
+  "cooking egg",
+  "cooking on campfire",
+  "cooking sausages",
+  "counting money",
+  "country line dancing",
+  "cracking neck",
+  "crawling baby",
+  "crossing river",
+  "crying",
+  "curling hair",
+  "cutting nails",
+  "cutting pineapple",
+  "cutting watermelon",
+  "dancing ballet",
+  "dancing charleston",
+  "dancing gangnam style",
+  "dancing macarena",
+  "deadlifting",
+  "decorating the christmas tree",
+  "digging",
+  "dining",
+  "disc golfing",
+  "diving cliff",
+  "dodgeball",
+  "doing aerobics",
+  "doing laundry",
+  "doing nails",
+  "drawing",
+  "dribbling basketball",
+  "drinking",
+  "drinking beer",
+  "drinking shots",
+  "driving car",
+  "driving tractor",
+  "drop kicking",
+  "drumming fingers",
+  "dunking basketball",
+  "dying hair",
+  "eating burger",
+  "eating cake",
+  "eating carrots",
+  "eating chips",
+  "eating doughnuts",
+  "eating hotdog",
+  "eating ice cream",
+  "eating spaghetti",
+  "eating watermelon",
+  "egg hunting",
+  "exercising arm",
+  "exercising with an exercise ball",
+  "extinguishing fire",
+  "faceplanting",
+  "feeding birds",
+  "feeding fish",
+  "feeding goats",
+  "filling eyebrows",
+  "finger snapping",
+  "fixing hair",
+  "flipping pancake",
+  "flying kite",
+  "folding clothes",
+  "folding napkins",
+  "folding paper",
+  "front raises",
+  "frying vegetables",
+  "garbage collecting",
+  "gargling",
+  "getting a haircut",
+  "getting a tattoo",
+  "giving or receiving award",
+  "golf chipping",
+  "golf driving",
+  "golf putting",
+  "grinding meat",
+  "grooming dog",
+  "grooming horse",
+  "gymnastics tumbling",
+  "hammer throw",
+  "headbanging",
+  "headbutting",
+  "high jump",
+  "high kick",
+  "hitting baseball",
+  "hockey stop",
+  "holding snake",
+  "hopscotch",
+  "hoverboarding",
+  "hugging",
+  "hula hooping",
+  "hurdling",
+  "hurling (sport)",
+  "ice climbing",
+  "ice fishing",
+  "ice skating",
+  "ironing",
+  "javelin throw",
+  "jetskiing",
+  "jogging",
+  "juggling balls",
+  "juggling fire",
+  "juggling soccer ball",
+  "jumping into pool",
+  "jumpstyle dancing",
+  "kicking field goal",
+  "kicking soccer ball",
+  "kissing",
+  "kitesurfing",
+  "knitting",
+  "krumping",
+  "laughing",
+  "laying bricks",
+  "long jump",
+  "lunge",
+  "making a cake",
+  "making a sandwich",
+  "making bed",
+  "making jewelry",
+  "making pizza",
+  "making snowman",
+  "making sushi",
+  "making tea",
+  "marching",
+  "massaging back",
+  "massaging feet",
+  "massaging legs",
+  "massaging person's head",
+  "milking cow",
+  "mopping floor",
+  "motorcycling",
+  "moving furniture",
+  "mowing lawn",
+  "news anchoring",
+  "opening bottle",
+  "opening present",
+  "paragliding",
+  "parasailing",
+  "parkour",
+  "passing American football (in game)",
+  "passing American football (not in game)",
+  "peeling apples",
+  "peeling potatoes",
+  "petting animal (not cat)",
+  "petting cat",
+  "picking fruit",
+  "planting trees",
+  "plastering",
+  "playing accordion",
+  "playing badminton",
+  "playing bagpipes",
+  "playing basketball",
+  "playing bass guitar",
+  "playing cards",
+  "playing cello",
+  "playing chess",
+  "playing clarinet",
+  "playing controller",
+  "playing cricket",
+  "playing cymbals",
+  "playing didgeridoo",
+  "playing drums",
+  "playing flute",
+  "playing guitar",
+  "playing harmonica",
+  "playing harp",
+  "playing ice hockey",
+  "playing keyboard",
+  "playing kickball",
+  "playing monopoly",
+  "playing organ",
+  "playing paintball",
+  "playing piano",
+  "playing poker",
+  "playing recorder",
+  "playing saxophone",
+  "playing squash or racquetball",
+  "playing tennis",
+  "playing trombone",
+  "playing trumpet",
+  "playing ukulele",
+  "playing violin",
+  "playing volleyball",
+  "playing xylophone",
+  "pole vault",
+  "presenting weather forecast",
+  "pull ups",
+  "pumping fist",
+  "pumping gas",
+  "punching bag",
+  "punching person (boxing)",
+  "push up",
+  "pushing car",
+  "pushing cart",
+  "pushing wheelchair",
+  "reading book",
+  "reading newspaper",
+  "recording music",
+  "riding a bike",
+  "riding camel",
+  "riding elephant",
+  "riding mechanical bull",
+  "riding mountain bike",
+  "riding mule",
+  "riding or walking with horse",
+  "riding scooter",
+  "riding unicycle",
+  "ripping paper",
+  "robot dancing",
+  "rock climbing",
+  "rock scissors paper",
+  "roller skating",
+  "running on treadmill",
+  "sailing",
+  "salsa dancing",
+  "sanding floor",
+  "scrambling eggs",
+  "scuba diving",
+  "setting table",
+  "shaking hands",
+  "shaking head",
+  "sharpening knives",
+  "sharpening pencil",
+  "shaving head",
+  "shaving legs",
+  "shearing sheep",
+  "shining shoes",
+  "shooting basketball",
+  "shooting goal (soccer)",
+  "shot put",
+  "shoveling snow",
+  "shredding paper",
+  "shuffling cards",
+  "side kick",
+  "sign language interpreting",
+  "singing",
+  "situp",
+  "skateboarding",
+  "ski jumping",
+  "skiing (not slalom or crosscountry)",
+  "skiing crosscountry",
+  "skiing slalom",
+  "skipping rope",
+  "skydiving",
+  "slacklining",
+  "slapping",
+  "sled dog racing",
+  "smoking",
+  "smoking hookah",
+  "snatch weight lifting",
+  "sneezing",
+  "sniffing",
+  "snorkeling",
+  "snowboarding",
+  "snowkiting",
+  "snowmobiling",
+  "somersaulting",
+  "spinning poi",
+  "spray painting",
+  "spraying",
+  "springboard diving",
+  "squat",
+  "sticking tongue out",
+  "stomping grapes",
+  "stretching arm",
+  "stretching leg",
+  "strumming guitar",
+  "surfing crowd",
+  "surfing water",
+  "sweeping floor",
+  "swimming backstroke",
+  "swimming breast stroke",
+  "swimming butterfly stroke",
+  "swing dancing",
+  "swinging legs",
+  "swinging on something",
+  "sword fighting",
+  "tai chi",
+  "taking a shower",
+  "tango dancing",
+  "tap dancing",
+  "tapping guitar",
+  "tapping pen",
+  "tasting beer",
+  "tasting food",
+  "testifying",
+  "texting",
+  "throwing axe",
+  "throwing ball",
+  "throwing discus",
+  "tickling",
+  "tobogganing",
+  "tossing coin",
+  "tossing salad",
+  "training dog",
+  "trapezing",
+  "trimming or shaving beard",
+  "trimming trees",
+  "triple jump",
+  "tying bow tie",
+  "tying knot (not on a tie)",
+  "tying tie",
+  "unboxing",
+  "unloading truck",
+  "using computer",
+  "using remote controller (not gaming)",
+  "using segway",
+  "vault",
+  "waiting in line",
+  "walking the dog",
+  "washing dishes",
+  "washing feet",
+  "washing hair",
+  "washing hands",
+  "water skiing",
+  "water sliding",
+  "watering plants",
+  "waxing back",
+  "waxing chest",
+  "waxing eyebrows",
+  "waxing legs",
+  "weaving basket",
+  "welding",
+  "whistling",
+  "windsurfing",
+  "wrapping present",
+  "wrestling",
+  "writing",
+  "yawning",
+  "yoga",
+  "zumba"
+]
\ No newline at end of file
diff --git a/test_data/kinetics/frame_extract.py b/test_data/kinetics/frame_extract.py
new file mode 100644
index 0000000..a583904
--- /dev/null
+++ b/test_data/kinetics/frame_extract.py
@@ -0,0 +1,14 @@
+import os
+
+basedir = '/home/acb11711tx/lzq/ModelVisualization/test_data/kinetics'
+video_names = [name[:-4] for name in os.listdir(basedir) if '.mp4' in name]
+for video_name in video_names:
+    video_name = video_name
+    savedir = os.path.join(basedir, video_name)
+    os.makedirs(savedir, exist_ok=True)
+
+    savedir_ = savedir.replace(' ', '\ ')
+    video_name_ = video_name.replace(' ', '\ ')
+    print(video_name_)
+    os.system(f'ffmpeg -i {video_name_}.mp4 -vf fps=10 {video_name_}/%09d.png')
+    print(f'{video_name} finished!')
\ No newline at end of file
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000001.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000001.png
new file mode 100644
index 0000000..b300937
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000001.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000002.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000002.png
new file mode 100644
index 0000000..361bcba
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000002.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000003.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000003.png
new file mode 100644
index 0000000..787899a
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000003.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000004.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000004.png
new file mode 100644
index 0000000..35ebb83
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000004.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000005.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000005.png
new file mode 100644
index 0000000..5cc80f9
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000005.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000006.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000006.png
new file mode 100644
index 0000000..e0590c3
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000006.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000007.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000007.png
new file mode 100644
index 0000000..a0f8978
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000007.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000008.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000008.png
new file mode 100644
index 0000000..2e27258
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000008.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000009.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000009.png
new file mode 100644
index 0000000..a4a6ed2
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000009.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000010.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000010.png
new file mode 100644
index 0000000..f61395b
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000010.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000011.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000011.png
new file mode 100644
index 0000000..3ecc8bc
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000011.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000012.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000012.png
new file mode 100644
index 0000000..fd6a993
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000012.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000013.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000013.png
new file mode 100644
index 0000000..051eebf
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000013.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000014.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000014.png
new file mode 100644
index 0000000..1577a50
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000014.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000015.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000015.png
new file mode 100644
index 0000000..8ad771e
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000015.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000016.png b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000016.png
new file mode 100644
index 0000000..e4e89d0
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-4lFmhj4x43c_66_10/000000016.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000001.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000001.png
new file mode 100644
index 0000000..5e95170
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000001.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000002.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000002.png
new file mode 100644
index 0000000..8b02c18
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000002.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000003.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000003.png
new file mode 100644
index 0000000..e6a44ee
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000003.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000004.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000004.png
new file mode 100644
index 0000000..ffa7203
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000004.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000005.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000005.png
new file mode 100644
index 0000000..82c6dee
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000005.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000006.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000006.png
new file mode 100644
index 0000000..6dadf78
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000006.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000007.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000007.png
new file mode 100644
index 0000000..91aa57c
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000007.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000008.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000008.png
new file mode 100644
index 0000000..233f995
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000008.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000009.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000009.png
new file mode 100644
index 0000000..5ca2cc7
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000009.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000010.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000010.png
new file mode 100644
index 0000000..582a41b
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000010.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000011.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000011.png
new file mode 100644
index 0000000..556269d
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000011.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000012.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000012.png
new file mode 100644
index 0000000..906b2b3
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000012.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000013.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000013.png
new file mode 100644
index 0000000..dd8302a
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000013.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000014.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000014.png
new file mode 100644
index 0000000..7e101d9
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000014.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000015.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000015.png
new file mode 100644
index 0000000..0f6f53d
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000015.png differ
diff --git a/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000016.png b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000016.png
new file mode 100644
index 0000000..0207c97
Binary files /dev/null and b/test_data/kinetics/sampled_frames/abseiling-BkIhp07ge9E_22_10/000000016.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000001.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000001.png
new file mode 100644
index 0000000..c4b91ee
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000001.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000002.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000002.png
new file mode 100644
index 0000000..80ad00e
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000002.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000003.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000003.png
new file mode 100644
index 0000000..6ee04af
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000003.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000004.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000004.png
new file mode 100644
index 0000000..929eab4
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000004.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000005.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000005.png
new file mode 100644
index 0000000..01e3dd9
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000005.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000006.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000006.png
new file mode 100644
index 0000000..efb85de
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000006.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000007.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000007.png
new file mode 100644
index 0000000..fcb719a
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000007.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000008.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000008.png
new file mode 100644
index 0000000..9654220
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000008.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000009.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000009.png
new file mode 100644
index 0000000..a91d3cd
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000009.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000010.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000010.png
new file mode 100644
index 0000000..5f7f021
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000010.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000011.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000011.png
new file mode 100644
index 0000000..1a3cd50
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000011.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000012.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000012.png
new file mode 100644
index 0000000..68decb5
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000012.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000013.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000013.png
new file mode 100644
index 0000000..0659dd9
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000013.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000014.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000014.png
new file mode 100644
index 0000000..176b977
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000014.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000015.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000015.png
new file mode 100644
index 0000000..01cc30e
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000015.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000016.png b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000016.png
new file mode 100644
index 0000000..01a0118
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-3sEk_w2T8Fw_15_10/000000016.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000001.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000001.png
new file mode 100644
index 0000000..3d712fa
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000001.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000002.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000002.png
new file mode 100644
index 0000000..fe6bd95
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000002.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000003.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000003.png
new file mode 100644
index 0000000..9138994
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000003.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000004.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000004.png
new file mode 100644
index 0000000..9c40a44
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000004.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000005.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000005.png
new file mode 100644
index 0000000..83356a8
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000005.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000006.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000006.png
new file mode 100644
index 0000000..5cff18c
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000006.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000007.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000007.png
new file mode 100644
index 0000000..cbe813f
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000007.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000008.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000008.png
new file mode 100644
index 0000000..bc162bb
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000008.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000009.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000009.png
new file mode 100644
index 0000000..197ced5
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000009.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000010.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000010.png
new file mode 100644
index 0000000..e5e3cee
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000010.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000011.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000011.png
new file mode 100644
index 0000000..11df013
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000011.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000012.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000012.png
new file mode 100644
index 0000000..f08d7f4
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000012.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000013.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000013.png
new file mode 100644
index 0000000..9ab5653
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000013.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000014.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000014.png
new file mode 100644
index 0000000..ce5300f
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000014.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000015.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000015.png
new file mode 100644
index 0000000..60455ca
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000015.png differ
diff --git a/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000016.png b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000016.png
new file mode 100644
index 0000000..f3ecdb2
Binary files /dev/null and b/test_data/kinetics/sampled_frames/ironing-K2COe93hgqM_96_10/000000016.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000001.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000001.png
new file mode 100644
index 0000000..e1d8866
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000001.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000002.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000002.png
new file mode 100644
index 0000000..566cb0b
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000002.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000003.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000003.png
new file mode 100644
index 0000000..3f109e0
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000003.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000004.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000004.png
new file mode 100644
index 0000000..a5e2582
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000004.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000005.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000005.png
new file mode 100644
index 0000000..bb92ea9
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000005.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000006.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000006.png
new file mode 100644
index 0000000..5fffc37
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000006.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000007.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000007.png
new file mode 100644
index 0000000..e27fe26
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000007.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000008.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000008.png
new file mode 100644
index 0000000..8732e93
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000008.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000009.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000009.png
new file mode 100644
index 0000000..30fe4fa
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000009.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000010.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000010.png
new file mode 100644
index 0000000..8f78ef0
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000010.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000011.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000011.png
new file mode 100644
index 0000000..46e7a11
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000011.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000012.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000012.png
new file mode 100644
index 0000000..b98c9c9
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000012.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000013.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000013.png
new file mode 100644
index 0000000..a9439a9
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000013.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000014.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000014.png
new file mode 100644
index 0000000..e43690e
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000014.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000015.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000015.png
new file mode 100644
index 0000000..7de505a
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000015.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000016.png b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000016.png
new file mode 100644
index 0000000..212bee8
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-JVAwfvQUlwA_46_10/000000016.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000001.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000001.png
new file mode 100644
index 0000000..256a67d
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000001.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000002.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000002.png
new file mode 100644
index 0000000..9a0896c
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000002.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000003.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000003.png
new file mode 100644
index 0000000..70135f6
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000003.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000004.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000004.png
new file mode 100644
index 0000000..aec1a90
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000004.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000005.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000005.png
new file mode 100644
index 0000000..3658850
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000005.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000006.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000006.png
new file mode 100644
index 0000000..4b43714
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000006.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000007.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000007.png
new file mode 100644
index 0000000..74db3a2
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000007.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000008.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000008.png
new file mode 100644
index 0000000..3c50d9c
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000008.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000009.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000009.png
new file mode 100644
index 0000000..d083fcf
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000009.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000010.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000010.png
new file mode 100644
index 0000000..b258ed7
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000010.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000011.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000011.png
new file mode 100644
index 0000000..cde84d1
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000011.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000012.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000012.png
new file mode 100644
index 0000000..f63547f
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000012.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000013.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000013.png
new file mode 100644
index 0000000..97266b3
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000013.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000014.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000014.png
new file mode 100644
index 0000000..bf07003
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000014.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000015.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000015.png
new file mode 100644
index 0000000..9c28e0f
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000015.png differ
diff --git a/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000016.png b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000016.png
new file mode 100644
index 0000000..6162d07
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing bass guitar-fUbFAXY-vxQ_2_10/000000016.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000001.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000001.png
new file mode 100644
index 0000000..1b00e4f
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000001.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000002.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000002.png
new file mode 100644
index 0000000..572ceea
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000002.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000003.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000003.png
new file mode 100644
index 0000000..8eaae80
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000003.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000004.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000004.png
new file mode 100644
index 0000000..aeebc95
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000004.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000005.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000005.png
new file mode 100644
index 0000000..1eafd89
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000005.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000006.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000006.png
new file mode 100644
index 0000000..19b95c0
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000006.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000007.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000007.png
new file mode 100644
index 0000000..cd2d8fb
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000007.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000008.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000008.png
new file mode 100644
index 0000000..deba515
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000008.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000009.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000009.png
new file mode 100644
index 0000000..e29576a
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000009.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000010.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000010.png
new file mode 100644
index 0000000..9894454
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000010.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000011.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000011.png
new file mode 100644
index 0000000..f1ee165
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000011.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000012.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000012.png
new file mode 100644
index 0000000..0d85753
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000012.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000013.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000013.png
new file mode 100644
index 0000000..f6bfed0
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000013.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000014.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000014.png
new file mode 100644
index 0000000..6c7557e
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000014.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000015.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000015.png
new file mode 100644
index 0000000..8500cef
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000015.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000016.png b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000016.png
new file mode 100644
index 0000000..2840447
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-2enXX9xrIp0_7_10/000000016.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000001.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000001.png
new file mode 100644
index 0000000..5b72dbd
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000001.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000002.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000002.png
new file mode 100644
index 0000000..d74a42b
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000002.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000003.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000003.png
new file mode 100644
index 0000000..abdf647
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000003.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000004.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000004.png
new file mode 100644
index 0000000..45b1bf0
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000004.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000005.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000005.png
new file mode 100644
index 0000000..e6cfbd8
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000005.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000006.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000006.png
new file mode 100644
index 0000000..4e797e4
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000006.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000007.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000007.png
new file mode 100644
index 0000000..b005e71
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000007.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000008.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000008.png
new file mode 100644
index 0000000..f88cc9a
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000008.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000009.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000009.png
new file mode 100644
index 0000000..439cb6a
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000009.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000010.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000010.png
new file mode 100644
index 0000000..5e02f9c
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000010.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000011.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000011.png
new file mode 100644
index 0000000..9ef2aa6
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000011.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000012.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000012.png
new file mode 100644
index 0000000..2a16429
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000012.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000013.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000013.png
new file mode 100644
index 0000000..f6eb754
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000013.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000014.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000014.png
new file mode 100644
index 0000000..43bc602
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000014.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000015.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000015.png
new file mode 100644
index 0000000..2d354a3
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000015.png differ
diff --git a/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000016.png b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000016.png
new file mode 100644
index 0000000..28859d8
Binary files /dev/null and b/test_data/kinetics/sampled_frames/playing guitar-wFWTW0q57ms_140_10/000000016.png differ
diff --git a/utils/CalAcc.py b/utils/CalAcc.py
new file mode 100644
index 0000000..070bf75
--- /dev/null
+++ b/utils/CalAcc.py
@@ -0,0 +1,72 @@
+import torch
+
+class AverageMeter(object):
+    """Computes and stores the average and current value"""
+
+    def __init__(self, name, fmt=':f'):
+        self.name = name
+        self.fmt = fmt
+        self.reset()
+
+    def reset(self):
+        self.val = 0
+        self.avg = 0
+        self.sum = 0
+        self.count = 0
+
+    def update(self, val, n=1):
+        self.val = val
+        self.sum += val * n
+        self.count += n
+        self.avg = self.sum / self.count
+
+    def __str__(self):
+        fmtstr = '{name} {val' + self.fmt + '} ({avg' + self.fmt + '})'
+        return fmtstr.format(**self.__dict__)
+
+# output: batch_size x num_class(1000)
+# target: batch_size x 1 (LongTensor)
+
+
+def accuracy(output, target, topk=(1, 5)):
+    """Computes the accuracy over the k top predictions for the specified values of k"""
+    with torch.no_grad():
+        maxk = max(topk)
+        batch_size = target.size(0)
+
+        _, pred = output.topk(maxk, 1, True, True)
+        pred = pred.t()
+        correct = pred.eq(target.view(1, -1).expand_as(pred))
+
+        res = []
+        for k in topk:
+            correct_k = correct[:k].view(-1).float().sum(0, keepdim=True)
+            res.append(correct_k.mul_(100.0 / batch_size))
+        return res
+
+def process_activations(activations, targets, softmaxed=True):
+    assert activations.shape[0] % targets.shape[0] == 0, \
+        f"Check the batch size of activations and targets!"
+
+    b_repeat = activations.shape[0] // targets.shape[0]
+    if b_repeat > 1:
+        targets = targets.repeat(b_repeat)
+
+    if not softmaxed:
+        soft_act = torch.nn.functional.softmax(activations, dim=1)
+    else:
+        soft_act = activations
+
+    row_idx = torch.arange(
+        soft_act.shape[0], dtype=torch.long, device=activations.device)
+    probs = soft_act[row_idx, targets]  # batch_size
+    pred_label_probs, pred_labels = torch.max(soft_act, dim=1)
+
+    return probs, pred_labels, pred_label_probs
+
+if __name__ == "__main__":
+    output = torch.tensor([[0.01, 0.1, 0.05, 0.9], [0.99, 0.4, 0.2, 0.3]])
+    target = torch.LongTensor([[3], [1]])
+
+    acc = accuracy(output, target, topk=(1, 3))
+    print(acc)
diff --git a/utils/ConfusionMatrix.py b/utils/ConfusionMatrix.py
new file mode 100644
index 0000000..a3f6d74
--- /dev/null
+++ b/utils/ConfusionMatrix.py
@@ -0,0 +1,69 @@
+import numpy as np
+# import matplotlib.pyplot as plt
+
+import matplotlib
+matplotlib.use("Agg")
+import matplotlib.pyplot as plt
+
+from sklearn import svm, datasets
+from sklearn.model_selection import train_test_split
+from sklearn.metrics import confusion_matrix
+from sklearn.utils.multiclass import unique_labels
+
+import pickle
+
+
+def plot_confusion_matrix(y_true, y_pred, classes,
+                          normalize=False,
+                          title=None,
+                          cmap=plt.cm.Blues):
+    """
+    This function prints and plots the confusion matrix.
+    Normalization can be applied by setting `normalize=True`.
+    """
+    if not title:
+        if normalize:
+            title = 'Normalized confusion matrix'
+        else:
+            title = 'Confusion matrix, without normalization'
+
+    # Compute confusion matrix
+    cm = confusion_matrix(y_true, y_pred)
+    # Only use the labels that appear in the data
+    classes = classes[unique_labels(y_true, y_pred)]
+    if normalize:
+        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]
+        print("Normalized confusion matrix")
+    else:
+        print('Confusion matrix, without normalization')
+
+    # print(cm)
+
+    fig, ax = plt.subplots(figsize=(10,10))
+    im = ax.imshow(cm, interpolation='nearest', cmap=plt.get_cmap('Blues'))
+    # ax.figure.colorbar(im, ax=ax)
+    plt.colorbar(im,fraction=0.046, pad=0.04)
+    # We want to show all ticks...
+    ax.set(xticks=np.arange(cm.shape[1]),
+           yticks=np.arange(cm.shape[0]),
+           # ... and label them with the respective list entries
+           xticklabels=classes, yticklabels=classes,
+           title=title,
+           ylabel='True label',
+           xlabel='Predicted label')
+
+    # Rotate the tick labels and set their alignment.
+    plt.setp(ax.get_xticklabels(), rotation=45, ha="right",
+             rotation_mode="anchor", fontsize=6)
+    plt.setp(ax.get_yticklabels(), fontsize=6)
+
+    # Loop over data dimensions and create text annotations.
+    # fmt = '.2f' if normalize else 'd'
+    # thresh = cm.max() / 2.
+    # for i in range(cm.shape[0]):
+    #     for j in range(cm.shape[1]):
+    #         ax.text(j, i, format(cm[i, j], fmt),
+    #                 ha="center", va="center",
+    #                 color="white" if cm[i, j] > thresh else "black")
+    fig.tight_layout()
+    return fig, ax
\ No newline at end of file
diff --git a/utils/ImageShow.py b/utils/ImageShow.py
new file mode 100644
index 0000000..8dec631
--- /dev/null
+++ b/utils/ImageShow.py
@@ -0,0 +1,268 @@
+import torch
+import torchvision.transforms.functional as TF
+
+import numpy as np
+from PIL import Image
+from skimage import transform, filters
+import math
+import os
+
+import matplotlib
+matplotlib.use("Agg")
+import matplotlib.pyplot as plt
+
+def pil_to_tensor(pil_image):
+    r"""Convert a PIL image to a tensor.
+    Args:
+        pil_image (:class:`PIL.Image`): PIL image.
+    Returns:
+        :class:`torch.Tensor`: the image as a :math:`3\times H\times W` tensor
+        in the [0, 1] range.
+    """
+    pil_image = np.array(pil_image)
+    if len(pil_image.shape) == 2:
+        pil_image = pil_image[:, :, None]
+    return torch.tensor(pil_image, dtype=torch.float32).permute(2, 0, 1) / 255
+
+def img_tensor_to_pil(img_tensor):
+    lim = [img_tensor.min(), img_tensor.max()]
+    img_tensor = img_tensor - lim[0]  # also makes a copy
+    img_tensor.mul_(1 / (lim[1] - lim[0]))
+    img_tensor = torch.clamp(img_tensor, min=0, max=1)
+    img_pil = TF.to_pil_image(img_tensor)
+    return img_pil
+
+def img_tensor_to_np(img_tensor):
+    lim = [img_tensor.min(), img_tensor.max()]
+    img_tensor = img_tensor - lim[0]  # also makes a copy
+    img_tensor.mul_(1 / (lim[1] - lim[0]))
+    img_tensor = torch.clamp(img_tensor, min=0, max=1)
+    img_np = img_tensor.numpy()
+    return img_np
+
+def voxel_tensor_to_np(voxel_tensor):
+    # voxel_tensor: CxTxHxW
+    voxel_np = []
+    for t in range(voxel_tensor.shape[1]):
+        img_np = img_tensor_to_np(voxel_tensor[:,t,:,:])
+        voxel_np.append(img_np)
+    voxel_np = np.stack(voxel_np, axis=1)
+    return voxel_np
+
+def map_to_colormap(attMap, resize=(), norm_map=False, blur=False):
+    attMap = attMap.copy()
+    if norm_map:
+        attMap -= attMap.min()
+        if attMap.max() > 0:
+            attMap /= attMap.max()
+
+    if resize != ():
+        attMap = transform.resize(attMap, resize, order=3)
+
+    if blur:
+        attMap = filters.gaussian(attMap, 0.02*max(resize))
+        attMap -= attMap.min()
+        attMap /= attMap.max()
+
+    cmap = plt.get_cmap('jet')
+    colormap = cmap(attMap)
+    colormap = np.delete(colormap, 3, 2)
+    colormap = np.transpose(colormap, (2,0,1))    #3x224x224
+    return attMap, colormap
+
+def overlap_map_on_img(img_np, attMap, norm_map=False, blur=False):
+    # img_np: CxHxW, attMap: h x w
+    plt.axis('off')
+    resized_map, colormap = map_to_colormap(attMap, resize=(img_np.shape[1:]), norm_map=False, blur=blur)
+    resized_map = 1*(1-resized_map**0.8)*img_np + (resized_map**0.8)*colormap
+    return resized_map
+
+def overlap_maps_on_voxel_np(voxel_np, attMaps, norm_map=False, blur=False):
+    overlaps = []
+    for t in range(voxel_np.shape[1]):
+        img_np = voxel_np[:,t,:,:]
+        attMap = attMaps[t,:,:]
+        overlap = overlap_map_on_img(img_np, attMap, norm_map, blur)
+        overlaps.append(overlap)
+    overlaps = np.stack(overlaps, axis=1)    # 3x16x224x224
+    return overlaps
+
+def img_np_show (img_np, interpolation='lanczos'):
+    bitmap = np.transpose(img_np, (1,2,0))  # HxWxC
+    handle = plt.imshow(
+        bitmap, interpolation=interpolation, vmin=0, vmax=1)
+    curr_ax = plt.gca()
+    curr_ax.axis('off')
+    return handle
+
+def imsc(img, *args, lim=None, quiet=False, interpolation='lanczos', **kwargs):
+    r"""Rescale and displays an image represented as a img.
+    The function scales the img :attr:`im` to the [0 ,1] range.
+    The img is assumed to have shape :math:`3\times H\times W` (RGB)
+    :math:`1\times H\times W` (grayscale).
+    Args:
+        img (:class:`torch.Tensor` or :class:`PIL.Image`): image.
+        quiet (bool, optional): if False, do not display image.
+            Default: ``False``.
+        lim (list, optional): maximum and minimum intensity value for
+            rescaling. Default: ``None``.
+        interpolation (str, optional): The interpolation mode to use with
+            :func:`matplotlib.pyplot.imshow` (e.g. ``'lanczos'`` or
+            ``'nearest'``). Default: ``'lanczos'``.
+    Returns:
+        :class:`torch.Tensor`: Rescaled image img.
+    """
+    if isinstance(img, Image.Image):
+        img = pil_to_tensor(img)
+    handle = None
+    with torch.no_grad():
+        if not lim:
+            lim = [img.min(), img.max()]
+        img = img - lim[0]  # also makes a copy
+        img.mul_(1 / (lim[1] - lim[0]))
+        img = torch.clamp(img, min=0, max=1)
+        if not quiet:
+            bitmap = img.expand(3, *img.shape[1:]).permute(1, 2, 0).cpu().numpy()
+            handle = plt.imshow(
+                bitmap, *args, interpolation=interpolation, **kwargs)
+            curr_ax = plt.gca()
+            curr_ax.axis('off')
+    return img, handle
+
+def plot_voxel(voxel, saliency, show_plot=False, save_path=None):
+    # batch_size = len(input)
+    num_frame = voxel.shape[1]
+    num_row = 2 * num_frame//8 
+
+    plt.clf()
+    fig = plt.figure(figsize=(16,num_row*2))
+    for i in range(num_frame):
+        plt.subplot(num_row, 8, (i//8)*16+i%8+1)
+        imsc(voxel[:,i,:,:])
+        plt.title(i, fontsize=8)
+
+        plt.subplot(num_row, 8, (i//8)*16+i%8+8+1)
+        imsc(saliency[:,i,:,:], interpolation='none')
+
+    # Save figure if path is specified.
+    if save_path:
+        save_dir = os.path.dirname(os.path.abspath(save_path))
+        # Create directory if necessary.
+        if not os.path.exists(save_dir):
+            os.makedirs(save_dir)
+        ext = os.path.splitext(save_path)[1].strip('.')
+        plt.savefig(save_path, format=ext, bbox_inches='tight')
+
+    # Show plot if desired.
+    if show_plot:
+        plt.show()
+
+def plot_voxel_wbbox(voxel, saliency, bbox_tensor, 
+                    show_plot=False, save_path=None):
+    # batch_size = len(input)
+    num_frame = voxel.shape[1]
+    num_row = 2 * num_frame//8 
+
+    for idx in range(num_frame):
+        x0, y0, x1, y1 = bbox_tensor[idx,:].tolist()
+        voxel[1, idx, y0:y1+1, x0] = 1.0
+        voxel[1, idx, y0:y1+1, x1] = 1.0
+        voxel[1, idx, y0, x0:x1+1] = 1.0
+        voxel[1, idx, y1, x0:x1+1] = 1.0
+
+    plt.clf()
+    fig = plt.figure(figsize=(16,num_row*2))
+    for i in range(num_frame):
+        plt.subplot(num_row, 8, (i//8)*16+i%8+1)
+        imsc(voxel[:,i,:,:])
+        plt.title(i, fontsize=8)
+
+        plt.subplot(num_row, 8, (i//8)*16+i%8+8+1)
+        imsc(saliency[:,i,:,:], interpolation='none')
+
+    # Save figure if path is specified.
+    if save_path:
+        save_dir = os.path.dirname(os.path.abspath(save_path))
+        # Create directory if necessary.
+        if not os.path.exists(save_dir):
+            os.makedirs(save_dir)
+        ext = os.path.splitext(save_path)[1].strip('.')
+        plt.savefig(save_path, format=ext, bbox_inches='tight')
+
+    # Show plot if desired.
+    if show_plot:
+        plt.show()
+
+def plot_voxel_np(voxel_np, saliency_np, title=None, 
+            show_plot=False, save_path=None):
+    # batch_size = len(input)
+    num_frame = voxel_np.shape[1]
+    num_row = 2 * num_frame//8 
+
+    plt.clf()
+    fig = plt.figure(figsize=(16,num_row*2))
+    for i in range(num_frame):
+        plt.subplot(num_row, 8, (i//8)*16+i%8+1)
+        img_np_show(voxel_np[:,i,:,:])
+        plt.title(i, fontsize=8)
+
+        plt.subplot(num_row, 8, (i//8)*16+i%8+8+1)
+        img_np_show(saliency_np[:,i,:,:], interpolation='none')
+    # fig.close()
+
+    if title is not None:
+        fig.suptitle(title, fontsize=14)
+
+    # Save figure if path is specified.
+    if save_path:
+        save_dir = os.path.dirname(os.path.abspath(save_path))
+        # Create directory if necessary.
+        if not os.path.exists(save_dir):
+            os.makedirs(save_dir)
+        ext = os.path.splitext(save_path)[1].strip('.')
+        plt.savefig(save_path, format=ext, bbox_inches='tight')
+
+    # Show plot if desired.
+    if show_plot:
+        plt.show()
+
+def plot_voxel_wbbox_np(voxel_np, saliency_np, bbox_tensor, title=None,
+                    show_plot=False, save_path=None):
+    # batch_size = len(input)
+    num_frame = voxel_np.shape[1]
+    num_row = 2 * num_frame//8 
+
+    for idx in range(num_frame):
+        x0, y0, x1, y1 = bbox_tensor[idx,:].tolist()
+        voxel_np[1, idx, y0:y1+1, x0] = 1.0
+        voxel_np[1, idx, y0:y1+1, x1] = 1.0
+        voxel_np[1, idx, y0, x0:x1+1] = 1.0
+        voxel_np[1, idx, y1, x0:x1+1] = 1.0
+
+    plt.clf()
+    fig = plt.figure(figsize=(16,num_row*2))
+    for i in range(num_frame):
+        plt.subplot(num_row, 8, (i//8)*16+i%8+1)
+        img_np_show(voxel_np[:,i,:,:])
+        plt.title(i, fontsize=8)
+
+        plt.subplot(num_row, 8, (i//8)*16+i%8+8+1)
+        img_np_show(saliency_np[:,i,:,:], interpolation='none')
+
+    if title is not None:
+        fig.suptitle(title, fontsize=14)
+
+    # Save figure if path is specified.
+    if save_path:
+        save_dir = os.path.dirname(os.path.abspath(save_path))
+        # Create directory if necessary.
+        if not os.path.exists(save_dir):
+            os.makedirs(save_dir)
+        ext = os.path.splitext(save_path)[1].strip('.')
+        plt.savefig(save_path, format=ext, bbox_inches='tight')
+
+    # Show plot if desired.
+    if show_plot:
+        plt.show()
+
+    plt.close(fig)
\ No newline at end of file
diff --git a/utils/LongRangeSample.py b/utils/LongRangeSample.py
new file mode 100644
index 0000000..812d9d2
--- /dev/null
+++ b/utils/LongRangeSample.py
@@ -0,0 +1,45 @@
+import math
+import random
+
+
+def long_range_rand(seq_len, num_seg):
+    r = int(seq_len / num_seg)
+    real_num_seg = int(math.ceil(seq_len / r))
+
+    frame_ind = []
+    for i in range(0, real_num_seg-1):
+        frame_ind.append(random.randint(i*r, (i+1)*r-1))
+    frame_ind.append(random.randint((real_num_seg-1)*r, seq_len-1))
+
+    frame_ind = frame_ind[len(frame_ind)-num_seg:]
+    return frame_ind
+
+def long_range_first(seq_len, num_seg):
+    r = int(seq_len / num_seg)
+
+    frame_ind = []
+    for i in range(0, seq_len, r):
+        frame_ind.append(i)
+    frame_ind = frame_ind[len(frame_ind)-num_seg:]
+    return frame_ind
+
+def long_range_last(seq_len, num_seg):
+    r = int(seq_len / num_seg)
+
+    frame_ind = []
+    for i in range(seq_len-1, -1, -r):
+        frame_ind.append(i)
+
+    frame_ind = frame_ind[0:num_seg]
+    frame_ind.reverse()
+    return frame_ind
+
+def long_range_sample(seq_len, num_seg, mode):
+    if mode == "random":
+        return long_range_rand(seq_len, num_seg)
+    elif mode == "first":
+        return long_range_first(seq_len, num_seg)
+    elif mode == "last":
+        return long_range_last(seq_len, num_seg)
+    else:
+        raise Exception(f"Given mode is unacceptable!")
diff --git a/visual_meth/grad_cam.py b/visual_meth/grad_cam.py
new file mode 100644
index 0000000..972928c
--- /dev/null
+++ b/visual_meth/grad_cam.py
@@ -0,0 +1,65 @@
+import torch
+import torch.nn.functional as F
+import numpy as np 
+
+# Backward hook
+observ_grad_ = []
+def backward_hook(m, i_grad, o_grad): 
+    global observ_grad_
+    observ_grad_.insert(0, o_grad[0].detach())
+
+# Forward hook
+observ_actv_ = []
+def forward_hook(m, i, o):
+    global observ_actv_
+    observ_actv_.append(o[0].detach())
+
+def grad_cam (inputs, labels, model, device, layer_name, norm_vis=True):
+    model.eval()   # Set model to evaluate mode
+    
+    bs, ch, nt, h, w = inputs.shape
+    assert ch == 3
+    assert labels.shape[0] == bs
+
+    # layer_dict = dict(model.module.named_children())
+    # assert layer_name in layer_dict, \
+    #     f'Given layer ({layer_name}) is not in model. {model}'
+    # observ_layer = layer_dict[layer_name]
+
+    observ_layer = model
+    for name in layer_name:
+        print(dict(observ_layer.named_children()).keys())
+        observ_layer = dict(observ_layer.named_children())[name]
+
+    observ_layer.register_backward_hook(backward_hook)
+    observ_layer.register_forward_hook(forward_hook)
+
+    inputs = inputs.to(device)
+    labels = labels.to(dtype=torch.long)
+
+    # Forward pass
+    outputs = model(inputs)
+    _, preds = torch.max(outputs, 1)
+
+    observ_actv = observ_actv_[0]   # 1 x C x num_f/8 x 56 x 56
+    observ_actv = torch.repeat_interleave(observ_actv, int(nt/observ_actv.shape[2]), dim=2)
+
+    # backward pass
+    backward_signals = torch.zeros_like(outputs, device=device)
+    for bidx in range(bs):
+        backward_signals[bidx, labels[bidx].cpu().item()] = 1.0
+    outputs.backward(backward_signals)
+
+    observ_grad = observ_grad_[0]   # 1 x C x num_f/8 x 56 x 56
+    observ_grad = torch.repeat_interleave(observ_grad, int(nt/observ_grad.shape[2]), dim=2)
+
+    observ_grad_w = observ_grad.mean(dim=4, keepdim=True).mean(dim=3, keepdim=True) # 1 x 512 x num_f x 1x1
+    out_masks = F.relu( (observ_grad_w*observ_actv).sum(dim=1, keepdim=True) ) # 1 x 1 x num_f x 14x14
+
+    if norm_vis:
+        out_masks = (out_masks - out_masks.min()) / (out_masks.max() - out_masks.min())
+
+    return out_masks
+
+
+
diff --git a/visual_meth/gradients.py b/visual_meth/gradients.py
new file mode 100644
index 0000000..e310de8
--- /dev/null
+++ b/visual_meth/gradients.py
@@ -0,0 +1,51 @@
+import torch
+import numpy as np 
+
+def gradients (inputs, labels, model, device, multiply_input=False,
+                 polarity='both', show_gray=False):
+    model.eval()   # Set model to evaluate mode
+
+    bs, ch, nt, h, w = inputs.shape
+    assert ch == 3
+    assert labels.shape[0] == bs
+
+    inputs = inputs.to(device)
+    # labels = labels.to(device)
+    labels = labels.to(dtype=torch.long)
+
+    # Get model outputs and calculate loss
+    inputs.requires_grad = True
+    with torch.set_grad_enabled(True):
+        outputs = model(inputs)
+        _, preds = torch.max(outputs, 1)
+
+        backward_signals = torch.zeros_like(outputs, device=device)
+        for bidx in range(bs):
+            backward_signals[bidx, labels[bidx].cpu().item()] = 1.0
+        outputs.backward(backward_signals)
+    inputs_grad = inputs.grad.cpu()
+    if multiply_input:
+        inputs_grad *= inputs.detach().cpu()
+
+    if polarity == 'positive':
+        inputs_grad = torch.clamp(inputs_grad, min=0.0)
+    elif polarity == 'negative':
+        inputs_grad = -1.0 * torch.clamp(inputs_grad, max=0.0)
+
+    if show_gray:
+        inputs_grad = inputs_grad.mean(dim=1, keepdim=True).repeat(1,ch,1,1,1)
+
+    grad_sorted = inputs_grad.view(-1).sort()[0]
+    length = grad_sorted.shape[0]
+    grad_min = 0.0
+    grad_max = grad_sorted[int(0.999 * length)]
+    grad_show = inputs_grad.clamp(min=grad_min, max=grad_max)
+    grad_show = (grad_show - grad_min) / (grad_max - grad_min)
+    # print(grad_show.max(), grad_show.min())
+
+    if polarity == 'positive':
+        grad_show[:,1:,...] = 0.0
+    elif polarity == 'negative':
+        grad_show[:,:2,...] = 0.0
+
+    return grad_show
diff --git a/visual_meth/integrated_grad.py b/visual_meth/integrated_grad.py
new file mode 100644
index 0000000..f6c3892
--- /dev/null
+++ b/visual_meth/integrated_grad.py
@@ -0,0 +1,57 @@
+import torch
+import numpy as np 
+
+def integrated_grad (inputs, labels, model, device, steps, 
+                        polarity='positive', show_gray=True):
+    model.eval()   # Set model to evaluate mode
+
+    bs, ch, nt, h, w = inputs.shape
+    assert ch == 3
+    assert labels.shape[0] == bs
+
+    inputs = inputs.to(device)
+    # labels = labels.to(device)
+    labels = labels.to(dtype=torch.long)
+
+    baseline = torch.zeros_like(inputs, device=device)
+    baseline_out = model(baseline)
+
+    backward_signals = torch.zeros_like(baseline_out, device=device)
+    for bidx in range(bs):
+        backward_signals[bidx, labels[bidx].cpu().item()] = 1.0
+
+    intg_grads = 0
+    for i in range(steps):
+        scaled_inputs = baseline + (float(i) / steps) * (inputs - baseline)
+        scaled_inputs.requires_grad = True
+
+        # Forward
+        outputs = model(scaled_inputs)
+        _, preds = torch.max(outputs, dim=1)
+
+        # Backward
+        outputs.backward(backward_signals)
+        intg_grads += scaled_inputs.grad.cpu() / steps
+    intg_grads *= (inputs - baseline).detach().cpu()
+        
+    if polarity == 'positive':
+        intg_grads = torch.clamp(intg_grads, min=0.0)
+    elif polarity == 'negative':
+        intg_grads = -1.0 * torch.clamp(intg_grads, max=0.0)
+
+    if show_gray:
+        intg_grads = intg_grads.mean(dim=1, keepdim=True).repeat(1,ch,1,1,1)
+
+    grad_sorted = intg_grads.view(-1).sort()[0]
+    length = grad_sorted.shape[0]
+    grad_min = 0.0
+    grad_max = grad_sorted[int(0.999 * length)]
+    grad_show = intg_grads.clamp(min=grad_min, max=grad_max)
+    grad_show = (grad_show - grad_min) / (grad_max - grad_min)  # NxCxTxHxW
+    
+    if polarity == 'positive':
+        intg_grads[:,1:,...] = 0.0
+    elif polarity == 'negative':
+        intg_grads[:,:2,...] = 0.0
+    return grad_show
+
diff --git a/visual_meth/perturbation.py b/visual_meth/perturbation.py
new file mode 100644
index 0000000..e1ee734
--- /dev/null
+++ b/visual_meth/perturbation.py
@@ -0,0 +1,414 @@
+import math
+import matplotlib.pyplot as plt
+import numpy as np
+
+import torch
+import torch.nn.functional as F
+import torch.optim as optim
+
+import sys
+sys.path.append(".")
+sys.path.append("..")
+from utils.CalAcc import process_activations
+from utils.ImageShow import *
+
+from torchray.utils import imsmooth, imsc
+from torchray.attribution.common import resize_saliency
+
+BLUR_PERTURBATION = "blur"
+FADE_PERTURBATION = "fade"
+
+PRESERVE_VARIANT = "preserve"
+DELETE_VARIANT = "delete"
+DUAL_VARIANT = "dual"
+
+def simple_log_reward(activation, target, variant):
+    N = target.shape[0]
+    bs = activation.shape[0]
+    b_repeat = int( bs // N )
+    device = activation.device
+
+    col_idx = target.repeat(b_repeat) # batch_size
+    row_idx = torch.arange(activation.shape[0], dtype=torch.long, device=device)   # batch_size
+    prob = activation[row_idx, col_idx] # batch_size
+
+    if variant == DELETE_VARIANT:
+        reward = -torch.log(1-prob)
+    elif variant == PRESERVE_VARIANT:
+        reward = -torch.log(prob)
+    elif variant == DUAL_VARIANT:
+        reward = (-torch.log(1-prob[N:])) + (-torch.log(prob[:N]))
+    else:
+        assert False
+    return reward
+
+class MaskGenerator:
+    def __init__(self, shape, step, sigma, batch_size=1, clamp=True, pooling_method='softmax'):
+        self.shape = shape
+        self.step = step
+        self.sigma = sigma
+        self.coldness = 20
+        self.batch_size = batch_size
+        self.clamp = clamp
+        self.pooling_method = pooling_method
+
+        assert int(step) == step
+
+        # self.kernel = lambda z: (z < 1).float()
+        self.kernel = lambda z: torch.exp(-2 * ((z - .5).clamp(min=0)**2))
+
+        self.margin = self.sigma
+        self.padding = 1 + math.ceil((self.margin + sigma) / step)
+        self.radius = 1 + math.ceil(sigma / step)
+        self.shape_in = [math.ceil(z / step) for z in self.shape]
+        self.shape_mid = [
+            z + 2 * self.padding - (2 * self.radius + 1) + 1
+            for z in self.shape_in
+        ]
+        self.shape_up = [self.step * z for z in self.shape_mid]
+        self.shape_out = [z - step + 1 for z in self.shape_up]
+
+        step_inv = [
+            torch.tensor(zm, dtype=torch.float32) /
+            torch.tensor(zo, dtype=torch.float32)
+            for zm, zo in zip(self.shape_mid, self.shape_up)
+        ]
+
+        # Generate kernel weights for smoothing mask_in
+        self.weight = torch.zeros((
+            1,
+            (2 * self.radius + 1)**2,
+            self.shape_out[0],
+            self.shape_out[1]
+        ))
+
+        for ky in range(2 * self.radius + 1):
+            for kx in range(2 * self.radius + 1):
+                uy, ux = torch.meshgrid(
+                    torch.arange(self.shape_out[0], dtype=torch.float32),
+                    torch.arange(self.shape_out[1], dtype=torch.float32)
+                )
+                iy = torch.floor(step_inv[0] * uy) + ky - self.padding
+                ix = torch.floor(step_inv[1] * ux) + kx - self.padding
+
+                delta = torch.sqrt(
+                    (uy - (self.margin + self.step * iy))**2 +
+                    (ux - (self.margin + self.step * ix))**2
+                )
+
+                k = ky * (2 * self.radius + 1) + kx
+
+                self.weight[0, k] = self.kernel(delta / sigma)
+
+    def generate(self, mask_in):
+        # mask_in: Nx1xHxW --> mask: Nx1xS_outxS_out
+        mask = F.unfold(mask_in,
+                        (2 * self.radius + 1,) * 2,
+                        padding=(self.padding,) * 2)
+        mask = mask.reshape(
+            mask_in.shape[0], -1, self.shape_mid[0], self.shape_mid[1])
+        mask = F.interpolate(mask, size=self.shape_up, mode='nearest')
+        mask = F.pad(mask, (0, -self.step + 1, 0, -self.step + 1))
+        mask = self.weight * mask
+
+        if self.pooling_method == 'sigmoid':
+            if self.coldness == float('+Inf'):
+                mask = (mask.sum(dim=1, keepdim=True) - 5 > 0).float()
+            else:
+                mask = torch.sigmoid(
+                    self.coldness * mask.sum(dim=1, keepdim=True) - 3
+                )
+        elif self.pooling_method == 'softmax':
+            if self.coldness == float('+Inf'):  # max normalization
+                mask = mask.max(dim=1, keepdim=True)[0]
+            else:   # smax normalization
+                mask = (
+                    mask * F.softmax(self.coldness * mask, dim=1)
+                ).sum(dim=1, keepdim=True)
+        elif self.pooling_method == 'sum':
+            mask = mask.sum(dim=1, keepdim=True)
+        else:
+            assert False, f"Unknown pooling method {self.pooling_method}"
+
+        m = round(self.margin)
+        if self.clamp:
+            mask = mask.clamp(min=0, max=1)
+        cropped = mask[:, :, m:m + self.shape[0], m:m + self.shape[1]]
+        return cropped, mask
+
+    def to(self, dev):
+        """Switch to another device.
+        Args:
+            dev: PyTorch device.
+        Returns:
+            MaskGenerator: self.
+        """
+        self.weight = self.weight.to(dev)
+        return self
+
+class Perturbation:
+    def __init__(self, input, num_levels=8, max_blur=20, type=BLUR_PERTURBATION):
+        self.type = type
+        self.num_levels = num_levels
+        self.pyramid = []
+        assert num_levels >= 2
+        assert max_blur > 0
+        with torch.no_grad():
+            for sigma in torch.linspace(0, 1, self.num_levels):
+                if type == BLUR_PERTURBATION:
+                    # input could be a batched tensor with size of NxCxHxW
+                    y = imsmooth(input, sigma=(1 - sigma) * max_blur)
+                    # ouput y has size of NxCxHxW
+                elif type == FADE_PERTURBATION:
+                    y = input * sigma
+                else:
+                    assert False
+                self.pyramid.append(y)
+            # self.pyramid = torch.cat(self.pyramid, dim=0)
+            self.pyramid = torch.stack(self.pyramid, dim=1) # NxLxCxHxW, L=num_levels
+
+    def apply(self, mask):
+        # mask: Nx1xHxW
+        n = mask.shape[0]
+        # starred expression: unpack a list to separated numbers
+        w = mask.reshape(n, 1, *mask.shape[1:]) # Nx1x1xHxW, mask.unsqueeze(1)
+        w = w * (self.num_levels - 1)   # w = 7*w
+        k = w.floor()   # Integral part of w
+        w = w - k       # Fractional part of w
+        k = k.long()    # Transfer k to long int
+
+        # y = self.pyramid[None, :] #1xLxCxHxW
+        # y = y.expand(n, *y.shape[1:]) #nxLxCxHxW
+        # k = k.expand(n, 1, *y.shape[2:])  #nx1xCxHxW
+        y = self.pyramid    # NxLxCxHxW
+        k = k.expand(n, 1, *y.shape[2:])    #Nx1xCxHxW
+        y0 = torch.gather(y, 1, k)  # select low level, Nx1xCxHxW
+        y1 = torch.gather(y, 1, torch.clamp(k + 1, max=self.num_levels - 1)) # select high level, Nx1xCxHxW
+
+        # return ((1 - w) * y0 + w * y1).squeeze(dim=1)
+        perturb_x = ((1 - w) * y0 + w * y1)    #Nx1xCxHxW 
+        return perturb_x
+
+    def to(self, dev):
+        """Switch to another device.
+        Args:
+            dev: PyTorch device.
+        Returns:
+            Perturbation: self.
+        """
+        self.pyramid.to(dev)
+        return self
+
+    def __str__(self):
+        return (
+            f"Perturbation:\n"
+            f"- type: {self.type}\n"
+            f"- num_levels: {self.num_levels}\n"
+            f"- pyramid shape: {list(self.pyramid.shape)}"
+        )
+
+def video_perturbation(model, input, target,
+                        areas=[0.1], method='spatiotemporal', perturb_type=FADE_PERTURBATION,
+                        max_iter=2000, num_levels=8, step=7, sigma=11, variant=PRESERVE_VARIANT,
+                        print_iter=None, debug=False, reward_func="simple_log", resize=False,
+                        resize_mode='bilinear', smooth=0, num_devices=1):
+    
+    if isinstance(areas, float):
+        areas = [areas]
+    momentum = 0.9
+    learning_rate = 0.05
+    regul_weight = 300
+    reward_weight = 100
+    device = input.device
+
+    iter_period = 2000
+
+    regul_weight_last = max(regul_weight / 2, 1)
+
+    # input shape: NxCxTxHxW (1x3x16x112x112)
+    batch_size = input.shape[0]
+    num_frame = input.shape[2]  #16
+    num_areas = len(areas)
+
+    if debug:
+        print(
+            f"spatiotemporal_perturbation:\n"
+            f"- target: {target}\n"
+            f"- areas: {areas}\n"
+            f"- variant: {variant}\n"
+            f"- max_iter: {max_iter}\n"
+            f"- step/sigma: {step}, {sigma}\n"
+            f"- voxel size: {list(input.shape)}\n"
+            f"- reward function: {reward_func}"
+        )
+    print(f"- Target: {target.detach().cpu().tolist()}")
+
+    # Disable gradients for model parameters.
+    for p in model.parameters():
+        p.requires_grad_(False)
+    model.eval()
+
+    ori_y = model(input)
+    ori_prob, ori_pred_label, _ = process_activations(ori_y, target, softmaxed=True)
+
+    # NxCxTxHxW --> N*T x CxHxW
+    pmt_inp = input.transpose(1,2).contiguous() # NxTxCxHxW
+    pmt_inp = pmt_inp.view(batch_size*num_frame, *pmt_inp.shape[2:])  # N*T x CxHxW
+
+    # Get the perturbation operator.
+    # perturbation.pyramid: T*N x LxCxHxW
+    perturbation = Perturbation(pmt_inp, num_levels=num_levels, 
+                                    type=perturb_type).to(device)
+
+    # Prepare the mask generator (generating mask(134x134) from pmask(16x16)).
+    shape = perturbation.pyramid.shape[3:]  # 112x112
+    mask_generator = MaskGenerator(shape, step, sigma, pooling_method='softmax').to(device)
+    h, w = mask_generator.shape_in  # h=112/step, w=112/step, 16x16
+    pmasks = torch.ones(batch_size*num_frame, 1, h, w).to(device)  #N*T x 1x16x16
+
+    max_area = np.prod(mask_generator.shape_out)
+    max_volume = np.prod(mask_generator.shape_out) * num_frame
+
+    # Prepare reference area vector.
+    if method == 'spatiotemporal':
+        vref = torch.ones(batch_size, max_volume).to(device)
+        vref[:, :int(max_volume * (1 - areas[0]))] = 0
+    elif method == 'extremal':
+        aref = torch.ones(batch_size, num_frame, max_area).to(device)
+        aref[:, :, :int(max_area * (1 - areas[0]))] = 0
+
+    # Initialize optimizer.
+    optimizer = optim.SGD([pmasks],
+                          lr=learning_rate,
+                          momentum=momentum,
+                          dampening=momentum)
+    hist = torch.zeros((batch_size, 2, 0))
+
+    for t in range(max_iter):
+        pmasks.requires_grad_(True)
+        masks, padded_masks = mask_generator.generate(pmasks)
+
+        if variant == DELETE_VARIANT:
+            perturb_x = perturbation.apply(1 - masks)  # N*T x 1xCxHxW
+        elif variant == PRESERVE_VARIANT:
+            perturb_x = perturbation.apply(masks)  # N*T x 1xCxHxW
+        elif variant == DUAL_VARIANT:
+            perturb_x = torch.cat((
+                perturbation.apply(masks),  #preserve
+                perturbation.apply(1 - masks),  #delete
+            ), dim = 1) # N*T x 2xCxHxW
+        else:
+            assert False
+
+        perturb_x = perturb_x.view(batch_size, num_frame, *perturb_x.shape[1:]) # NxTx2xCxHxW
+        perturb_x = perturb_x.permute(2, 0, 3, 1, 4, 5).contiguous()   # 2xNxCxTxHxW
+        perturb_x = perturb_x.view(perturb_x.shape[0]*perturb_x.shape[1], *perturb_x.shape[2:])    # 2*N x CxTxHxW
+
+        masks = masks.view(batch_size, num_frame, *masks.shape[1:]).transpose(1,2) # Nx1xTxHxW
+        padded_masks = padded_masks.view(batch_size, num_frame, \
+                                *padded_masks.shape[1:]).transpose(1,2)    # Nx1xTx S_out x S_out
+
+        # Evaluate the model on the masked data
+        # The input of model should have size of NxCxTxHxW
+        y = model(perturb_x)    # 2*N x num_classes
+        y = F.softmax(y, dim=1)
+
+        # Cal probability
+        prob, pred_label, pred_label_prob = process_activations(y, target, softmaxed=True)
+
+        # Get reward.
+        if reward_func == "simple_log":
+            reward = simple_log_reward(y, target, variant=variant)
+        else:
+            raise Exception("Do not support other reward function now.")
+        # print(f"Reward shape: {reward.shape}")
+        reward = reward.view(batch_size, -1).mean(dim=1) * reward_weight  # batch_size
+
+        # Area regularization.
+        # padded_masks: N x 1 x T x S_out x S_out
+        if method == 'spatiotemporal':
+            mask_sorted = padded_masks.squeeze(1).reshape(batch_size, -1).sort(dim=1)[0]
+            regul = ((mask_sorted - vref)**2).mean(dim=1) * regul_weight # batch_size
+        elif method == 'extremal':
+            mask_sorted = padded_masks.squeeze(1).reshape(batch_size, num_frame, -1).sort(dim=2)[0]
+            regul = ((mask_sorted - aref)**2).mean(dim=2).mean(dim=1) * regul_weight # batch_size
+
+        # Energy summary
+        energy = (reward + regul).sum() 
+
+        # Gradient step.
+        optimizer.zero_grad()
+        energy.backward()
+        optimizer.step()
+
+        pmasks.data = pmasks.data.clamp(0, 1)
+
+        # Record energy
+        # hist: batch_size x 2 x num_iter
+        hist_item = torch.cat(
+                    (
+                        reward.detach().cpu().view(-1, 1, 1),
+                        regul.detach().cpu().view(-1, 1, 1)
+                    ), dim=1)
+        # print(f"Item shape: {hist_item.shape}")
+        hist = torch.cat((hist,hist_item), dim=2)
+
+        # Diagnostics.
+        # debug_this_iter = debug and (regul_weight / regul_weight_last >= 2)
+
+        # if (print_iter is not None and t % print_iter == 0) or debug_this_iter:
+        if (print_iter != None) and (t % print_iter == 0):
+            print("[{:04d}/{:04d}]".format(t + 1, max_iter), end="\n")
+            for i in range(batch_size):
+                if variant == "dual":
+                    print(" [area:{:.2f} loss:{:.2f} reg:{:.2f} presv:{:.2f}/{} del:{:.2f}/{}]".format(
+                        areas[0], hist[i, 0, -1], hist[i, 1, -1], 
+                        prob[i], pred_label[i], 
+                        prob[i+batch_size], pred_label[i+batch_size]), end="")
+                else:
+                    print(" [area:{:.2f} loss:{:.2f} reg:{:.2f} {}:{:.2f}/{}]".format(
+                        areas[0], hist[i, 0, -1], hist[i, 1, -1],
+                        variant, prob[i], pred_label[i]), end="")
+                print()
+
+        # if debug_this_iter:
+        #     regul_weight_last = regul_weight
+        #     # for i, a in enumerate(areas):
+        #     for i in range(batch_size):
+        #         plt.figure(i, figsize=(20, 6))
+        #         plt.clf()
+        #         ncols = 4 if variant == DUAL_VARIANT else 3
+        #         plt.subplot(1, ncols, 1)
+        #         plt.plot(hist[i, 0].numpy())
+        #         plt.plot(hist[i, 1].numpy())
+        #         plt.plot(hist[i].sum(dim=0).numpy())
+        #         plt.legend(('energy', 'regul', 'both'))
+        #         plt.title(f'target area:{areas[0]:.2f}')
+        #         mask = padded_masks[i,:,7,:,:]
+        #         plt.subplot(1, ncols, 2)
+        #         imsc(mask, lim=[0, 1])
+        #         plt.title(
+        #             f"min:{mask.min().item():.2f}"
+        #             f" max:{mask.max().item():.2f}"
+        #             f" area:{mask.sum() / mask.numel():.2f}")
+        #         plt.subplot(1, ncols, 3)
+        #         imsc(perturb_x[i,:,7,:,:])
+        #         if variant == DUAL_VARIANT:
+        #             plt.subplot(1, ncols, 4)
+        #             imsc(perturb_x[i + batch_size,:,7,:,:])
+        #         plt.pause(0.001)
+
+    masks = masks.detach()
+    # Resize saliency map.
+    list_mask = []
+    for frame_idx in range(num_frame):
+        mask = masks[:,:,frame_idx,:,:]
+        mask = resize_saliency(pmt_inp, mask, resize, mode=resize_mode)
+        # Smooth saliency map.
+        if smooth > 0:
+            mask = imsmooth(mask, sigma=smooth * min(mask.shape[2:]), padding_mode='constant')
+        list_mask.append(mask)
+    masks = torch.stack(list_mask, dim=2)   # NxCxTxHxW
+
+    # masks: NxCxTxHxW; hist: Nx3xmax_iter
+    return masks, hist, perturb_x
\ No newline at end of file