YOWOv2: A Stronger yet Efficient Multi-level Detection Framework for Real-time Spatio-temporal Action Detection
- We recommend you to use Anaconda to create a conda environment:
conda create -n trash python=3.9
- Then, activate the environment:
conda activate trash
- Requirements:
pip install -r requirements.txt
trash __videos __ Walking __video1.mp4
| | |__video2.mp4
| | |__.....
| |__ trashDumping __video1.mp4
| |__video2.mp4
| |__.....
|__label __ Walking __video1 __ frame11.txt
| | | |__ frame12.txt
| | | |__ .....
| | |__video2 __ frame11.txt
| | | |__ frame12.txt
| | | |__ .....
| | |__ .....
| |__ trashDumping __video1 __ frame21.txt
| | |__ frame22.txt
| | |__ .....
| |__video2 __ frame21.txt
| | |__ frame22.txt
| | |__ .....
| |__ .....
|__rgb-images __ Walking __video1 __ frame1.jpg
| | | |__ frame2.jpg
| | | |__ .....
| | |__video2 __ frame1.jpg
| | | |__ frame2.jpg
| | | |__ .....
| | |__ .....
| |__ trashDumping __video1 __ frame1.jpg
| | |__ frame2.jpg
| | |__ .....
| |__video2 __ frame1.jpg
| | |__ frame2.jpg
| | |__ .....
| |__ .....
|__testlist.txt
|__trainlist.txt
You can download Trash Dumping from the following links:
- Create folder labels, rgb-images
python extract_frame/video2frame.py --folder_videos trash/videos/Walking --label Walking
python extract_frame/video2frame.py --folder_videos trash/videos/trashDumping --label trashDumping
- Create testlist.txt, trainlist.txt
python trash/build_split.py
- Trash Dumping
Model | weight |
---|---|
YOWOv2-Nano | ckpt - chua co |
YOWOv2-Tiny | ckpt - chua co |
YOWOv2-medium | ckpt |
- Trash Dumping
For example:
python train.py --cuda -d trash --root . -v yowo_v2_nano --num_workers 2 --num_classes 2 --eval_epoch 1 --max_epoch 8 --lr_epoch 2 3 4 5 -lr 0.0001 -ldr 0.5 -bs 8 -accu 16 -K 16
or you can just run the script:
sh train_trash.sh
# run demo
python demo.py --cuda \
-v yowo_v2_medium \
--num_classes 2 \
-size 224 \
--weight checkpoints/trash/yowo_v2_medium/yowo_v2_medium_epoch_50.pth \
--video video_test/v_Basketball_g01_c02.mp4 \
--vis_thresh 0.5 \
-d trash \
or you can just run the script:
bash demo_trash.sh
Author : Tran Phuong Nam
Contact : [email protected]
Mentor : Cong Tran
Lab : Naver
@article{yang2023yowov2,
title={YOWOv2: A Stronger yet Efficient Multi-level Detection Framework for Real-time Spatio-temporal Action Detection},
author={Yang, Jianhua and Kun, Dai},
journal={arXiv preprint arXiv:2302.06848},
year={2023}
}