forked from jchenghu/ExpansionNet_v2
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdata_generator.py
134 lines (108 loc) · 6.05 KB
/
data_generator.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
import h5py
import numpy as np
from PIL import Image as PIL_Image
import torchvision
import torch
import argparse
from argparse import Namespace
from torch.nn.parameter import Parameter
from time import time
from data.coco_dataset import CocoDatasetKarpathy
torch.autograd.set_detect_anomaly(False)
torch.set_num_threads(1)
torch.set_num_interop_threads(1)
import functools
print = functools.partial(print, flush=True)
DEFAULT_RANK = 0
def convert_time_as_hhmmss(ticks):
return str(int(ticks / 60)) + " m " + \
str(int(ticks) % 60) + " s"
def generate_data(path_args):
coco_dataset = CocoDatasetKarpathy(images_path=path_args.images_path,
coco_annotations_path=args.captions_path + "dataset_coco.json",
train2014_bboxes_path=args.captions_path + "train2014_instances.json",
val2014_bboxes_path=args.captions_path + "val2014_instances.json",
preproc_images_hdf5_filepath=None,
precalc_features_hdf5_filepath=None,
limited_num_train_images=None,
limited_num_val_images=5000)
from models.swin_transformer_mod import SwinTransformer
model = SwinTransformer(
img_size=384, patch_size=4, in_chans=3,
embed_dim=192, depths=[2, 2, 18, 2], num_heads=[6, 12, 24, 48],
window_size=12, mlp_ratio=4., qkv_bias=True, qk_scale=None,
drop_rate=0.0, attn_drop_rate=0.0,
drop_path_rate=0.0,
norm_layer=torch.nn.LayerNorm, ape=False, patch_norm=True,
use_checkpoint=False)
def load_backbone_only_from_save(model, state_dict, prefix=None):
own_state = model.state_dict()
for name, param in state_dict.items():
if prefix is not None and name.startswith(prefix):
name = name[len(prefix):]
if name not in own_state:
print("Not found: " + str(name))
continue
if isinstance(param, Parameter):
param = param.data
own_state[name].copy_(param)
print("Found: " + str(name))
save_model_path = path_args.save_model_path
map_location = {'cuda:%d' % DEFAULT_RANK: 'cuda:%d' % DEFAULT_RANK}
checkpoint = torch.load(save_model_path, map_location=map_location)
if 'model_state_dict' in checkpoint.keys():
print("Custom save point found")
load_backbone_only_from_save(model, checkpoint['model_state_dict'], prefix='swin_transf.')
else:
print("Custom save point not found")
load_backbone_only_from_save(model, checkpoint['model'], prefix=None)
print("Loading phase ended")
model = model.to(DEFAULT_RANK)
test_preprocess_layers_1 = [torchvision.transforms.Resize((384, 384))]
test_preprocess_layers_2 = [torchvision.transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])]
test_preprocess_1 = torchvision.transforms.Compose(test_preprocess_layers_1)
test_preprocess_2 = torchvision.transforms.Compose(test_preprocess_layers_2)
model.eval()
with torch.no_grad():
hdf5_file = h5py.File(path_args.output_path, 'w')
def apply_model(model, file_path):
pil_image = PIL_Image.open(file_path)
if pil_image.mode != 'RGB':
pil_image = PIL_Image.new("RGB", pil_image.size)
preprocess_pil_image = test_preprocess_1(pil_image)
tens_image = torchvision.transforms.ToTensor()(preprocess_pil_image)
tens_image = test_preprocess_2(tens_image).to(DEFAULT_RANK)
output = model(tens_image.unsqueeze(0))
return output.squeeze(0)
for i in range(coco_dataset.train_num_images):
img_path, img_id = coco_dataset.get_image_path(coco_dataset.train_num_images - i - 1,
CocoDatasetKarpathy.TrainSet_ID)
output = apply_model(model, img_path)
hdf5_file.create_dataset(str(img_id) + '_features', data=np.array(output.cpu()))
if (i+1) % 5000 == 0 or (i+1) == coco_dataset.train_num_images:
print("Train " + str(i+1) + " / " + str(coco_dataset.train_num_images) + " completed")
for i in range(coco_dataset.test_num_images):
img_path, img_id = coco_dataset.get_image_path(i, CocoDatasetKarpathy.TestSet_ID)
output = apply_model(model, img_path)
hdf5_file.create_dataset(str(img_id) + '_features', data=np.array(output.cpu()))
if (i+1) % 2500 == 0 or (i+1) == coco_dataset.test_num_images:
print("Test " + str(i+1) + " / " + str(coco_dataset.test_num_images) + " completed")
for i in range(coco_dataset.val_num_images):
img_path, img_id = coco_dataset.get_image_path(i, CocoDatasetKarpathy.ValidationSet_ID)
output = apply_model(model, img_path)
hdf5_file.create_dataset(str(img_id) + '_features', data=np.array(output.cpu()))
if (i+1) % 2500 == 0 or (i+1) == coco_dataset.test_num_images:
print("Val " + str(i+1) + " / " + str(coco_dataset.test_num_images) + " completed")
print("[GPU: " + str(DEFAULT_RANK) + " ] Closing...")
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='Image Captioning')
parser.add_argument('--save_model_path', type=str, default='./github_ignore_material/saves/')
parser.add_argument('--output_path', type=str, default='./github_ignore_material/raw_data/precalc_features.hdf5')
parser.add_argument('--images_path', type=str, default='/tmp/images/')
parser.add_argument('--captions_path', type=str, default='./github_ignore_material/raw_data/')
args = parser.parse_args()
path_args = Namespace(save_model_path=args.save_model_path,
output_path=args.output_path,
images_path=args.images_path,
captions_path=args.captions_path)
generate_data(path_args=path_args)