SSD300 For PyTorch

This repository provides a script and recipe to train the SSD300 model to achieve state of the art accuracy. The codes of model architecture comes from NVIDIA(NVIDIA SSD pytorch), and the method belongs to paper SSD: Single Shot MultiBox Detector .

Feature support matrix

Copy from NVIDIA SSD pytorch

The following features are supported by this model.

Feature	SSD300 PyTorch
AMP	Yes
APEX DDP	Yes
NVIDIA DALI	No

Features

AMP is an abbreviation used for automatic mixed precision training.

DDP stands for DistributedDataParallel and is used for multi-GPU training.

NVIDIA DALI - DALI is a library accelerating data preparation pipeline. To accelerate your input pipeline, you only need to define your data loader with the DALI library. For details, see example sources in this repo or see the DALI documentation

Usage

1. Clone the repository.

git clone https://gitee.com/yann_qi/ssd-for-pytorch.git

or

git clone https://github.com/yannqi/SSD-for-pytorch.git

2. Download and preprocess the dataset.

   The SSD model was trained on the COCO 2017 dataset. You can download the dataset on COCO Download.

   NOTE: Make the dataset root like below:

    └──  COCO 
        ├──images
            ├── train2017: All train images(118287 images)
            ├── val2017: All validate images(5000 images)
        ├── annotations
            ├── instances_train2017.json
            ├── instances_val2017.json
            ├── captions_train2017.json
            ├── captions_val2017.json
            ├── person_keypoints_train2017.json
            └── person_keypoints_val2017.json
        └── coco_labels.txt
   

3. Config Setting

   Set the config in the data/coco.yaml

4. Train the model.(Unnecessary, you can download the pretrained checkpoint.)

   • Single GPU

   sh scripts/single_gpu.sh

   • Multi GPU

   sh scripts/multi_gpu.sh

5. Evaluate the model on the COCO dataset.

   Just run the python test.py

Checkpoint Download

You can download the Checkpoint in the https://catalog.ngc.nvidia.com/models.

And I also put them in the Google Drive, you can download them from https://drive.google.com/drive/folders/1ohDQPiR14-RKpC0dc2KtpFirZWffw45A?usp=sharing (Forgive me because of BaiDu Yun is too slow.)

Results

 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.250
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.424
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.255
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.074
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.268
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.400
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.237
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.344
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.359
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.116
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.392
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.550

Model FLOPs: 20.213G Params: 22.895M
Model summary: 173 layers, 22894902 parameters, 22894902 gradients

Data preprocessing

Copy from NVIDIA SSD pytorch

Before we feed data to the model, both during training and inference, we perform:

• JPEG decoding
• normalization with a mean = [0.485, 0.456, 0.406] and std dev = [0.229, 0.224, 0.225]
• encoding bounding boxes
• resizing to 300x300

Additionally, during training, data is:

• randomly shuffled
• samples without annotations are skipped

Data augmentation

During training we perform the following augmentation techniques:

• Random crop using the algorithm described in the SSD: Single Shot MultiBox Detector paper
• Random horizontal flip
• Color jitter