wangxiang1230 / OadTR

Code for our ICCV 2021 Paper "OadTR: Online Action Detection with Transformers".

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

OadTR

Code for our ICCV2021 paper: "OadTR: Online Action Detection with Transformers" ["Paper"]

Update

  • July 28, 2021: Our Paper "OadTR: Online Action Detection with Transformers" was accepted by ICCV2021. At the same time, we released THUMOS14-Kinetics feature.

Dependencies

  • pytorch==1.6.0
  • json
  • numpy
  • tensorboard-logger
  • torchvision==0.7.0

Prepare

  • Unzip the anno file "./data/anno_thumos.zip"
  • Download the feature THUMOS14-Anet feature (Note: HDD and TVSeries are available by contacting the authors of the datasets and signing agreements due to the copyrights. You can use this Repo to extract features.)

Training

python main.py --num_layers 3 --decoder_layers 5 --enc_layers 64 --output_dir models/en_3_decoder_5_lr_drop_1

Validation

python main.py --num_layers 3 --decoder_layers 5 --enc_layers 64 --output_dir models/en_3_decoder_5_lr_drop_1 --eval --resume models/en_3_decoder_5_lr_drop_1/checkpoint000{}.pth

Citing OadTR

Please cite our paper in your publications if it helps your research:

@inproceedings{wang2021oadtr,
  title={Oadtr: Online action detection with transformers},
  author={Wang, Xiang and Zhang, Shiwei and Qing, Zhiwu and Shao, Yuanjie and Zuo, Zhengrong and Gao, Changxin and Sang, Nong},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={7565--7575},
  year={2021}
}

About

Code for our ICCV 2021 Paper "OadTR: Online Action Detection with Transformers".

License:MIT License


Languages

Language:Python 100.0%