A pytorch easy re-implement of "YOLOX: Exceeding YOLO Series in 2021"

1. Notes

This is a pytorch easy re-implement of "YOLOX: Exceeding YOLO Series in 2021" [https://arxiv.org/abs/2107.08430]
The repo is still under development

2. Environment

pytorch>=1.7.0, python>=3.6, Ubuntu/Windows, see more in 'requirements.txt'

cd /path/to/your/work
git clone https://github.com/zhangming8/yolox-pytorch.git
cd yolox-pytorch
download pre-train weights in Model Zoo to /path/to/your/work/weights

3. Object Detection

Model Zoo

All weights can be downloaded from GoogleDrive or BaiduDrive(code:bc72)

Model	test size	mAPval 0.5:0.95	mAPtest 0.5:0.95	Params (M)
yolox-nano	416	25.4	25.7	0.91
yolox-tiny	416	33.1	33.2	5.06
yolox-s	640	39.3	39.6	9.0
yolox-m	640	46.2	46.4	25.3
yolox-l	640	49.5	50.0	54.2
yolox-x	640	50.5	51.1	99.1
yolox-x	800	51.2	51.9	99.1

The weights were converted from YOLOX. mAP was reevaluated on COCO val2017 and test2017, and some results are slightly better than the official implement. You can reproduce them by scripts in 'evaluate.sh'

Dataset

download COCO:
http://images.cocodataset.org/zips/train2017.zip
http://images.cocodataset.org/zips/val2017.zip
http://images.cocodataset.org/annotations/annotations_trainval2017.zip

unzip and put COCO dataset in following folders:
/path/to/dataset/annotations/instances_train2017.json
/path/to/dataset/annotations/instances_val2017.json
/path/to/dataset/images/train2017/*.jpg
/path/to/dataset/images/val2017/*.jpg

change opt.dataset_path = "/path/to/dataset" in 'config.py'

Train

a. Train from scratch:(backbone="CSPDarknet-s" means using yolox-s, and you can change it to any other backbone, eg: CSPDarknet-nano, tiny, s, m, l, x)
python train.py gpus='0' backbone="CSPDarknet-s" num_epochs=300 exp_id="coco_CSPDarknet-s_640x640" use_amp=False val_intervals=1 data_num_workers=8

b. Finetune, download pre-trained weight on COCO and finetune on customer dataset:
python train.py gpus='0' backbone="CSPDarknet-s" num_epochs=300 exp_id="coco_CSPDarknet-s_640x640" use_amp=False val_intervals=1 data_num_workers=8 load_model="../weights/yolox-s.pth" resume=False

c. Resume, you can use 'resume=True' when your training is accidentally stopped:
python train.py gpus='0' backbone="CSPDarknet-s" num_epochs=300 exp_id="coco_CSPDarknet-s_640x640" use_amp=False val_intervals=1 data_num_workers=8 load_model="exp/coco_CSPDarknet-s_640x640/model_last.pth" resume=True

d. Some tips:
Ⅰ You can also change params in 'train.sh'(these params will replace opt.xxx in config.py) and use 'sh train.sh' to train
Ⅱ if you want to close mulit-size training, change opt.random_size = None or (20, 21) in 'config.py')
Ⅲ mulit-gpu train: change opt.gpus = "3,5,6,7"

Evaluate

The trained weights will be saved in './exp/your_exp_id/model_xx.pth'
change 'load_model'='weight/path/to/evaluate.pth' and backbone='backbone-type' in 'evaluate.sh'
sh evaluate.sh

Predict/Inference/Demo

a. Predict images, change img_dir and load_model
python predict.py gpus='0' backbone="CSPDarknet-s" vis_thresh=0.3 load_model="exp/coco_CSPDarknet-s_640x640/model_best.pth" img_dir='/path/to/dataset/images/val2017'

b. Predict video
python predict.py gpus='0' backbone="CSPDarknet-s" vis_thresh=0.3 load_model="exp/coco_CSPDarknet-s_640x640/model_best.pth" video_dir='/path/to/your/video.mp4'

You can also change params in 'predict.sh', and use 'sh predict.sh'

Train Customer Dataset(VOC format)

1. put your annotations(.xml) and images(.jpg) into:
    /path/to/voc_data/images/train2017/*.jpg  # train images
    /path/to/voc_data/images/train2017/*.xml  # train xml annotations
    /path/to/voc_data/images/val2017/*.jpg  # val images
    /path/to/voc_data/images/val2017/*.xml  # val xml annotations

2. change opt.label_name = ['your', 'dataset', 'label'] in 'config.py'
   change opt.dataset_path = '/path/to/voc_data' in 'config.py'

3. python tools/voc_to_coco.py
   Converted COCO format annotation will be saved into:
    /path/to/voc_data/annotations/instances_train2017.json
    /path/to/voc_data/annotations/instances_val2017.json

4. (Optional) you can visualize the converted annotations by:
    python tools/show_coco_anns.py
    Here is an analysis of the COCO annotation https://blog.csdn.net/u010397980/article/details/90341223?spm=1001.2014.3001.5501

5. run train.sh, evaluate.sh, predict.sh (are the same as COCO)

4. Multi/One-class Multi-object Tracking(MOT)

one-class/single-class MOT Dataset

DOING

Multi-class MOT Dataset

DOING, not ready
1. download and unzip VisDrone dataset http://aiskyeye.com/download/multi-object-tracking_2021

2. put train and val dataset into:
/path/to/dataset/VisDrone/VisDrone2019-MOT-train  # This folder contains two subfolders, 'annotations' and 'sequences'
/path/to/dataset/VisDrone/VisDrone2019-MOT-val  # This folder contains two subfolders, 'annotations' and 'sequences'

3. change opt.dataset_path = "/path/to/dataset/VisDrone" in 'config.py'
4. python tools/visdrone_mot_to_coco.py  # converted to COCO format
5.(Optional) python tools/show_coco_anns.py  # visualized tracking id

6. set class name and tracking id number
change opt.label_name=['pedestrian', 'people', 'bicycle', 'car', 'van', 'truck', 'tricycle', 'awning-tricycle', 'bus', 'motor']
change opt.tracking_id_nums=[1829, 853, 323, 3017, 295, 159, 215, 79, 55, 749]
change opt.reid_dim=128

Train

DOING

Evaluate

DOING

Predict/Inference/Demo

DOING

5. Acknowledgement

https://github.com/Megvii-BaseDetection/YOLOX
https://github.com/PaddlePaddle/PaddleDetection
https://github.com/open-mmlab/mmdetection