FastGait

The official implementation for the Dynamic Aggregated Network for Gait Recognition which is accepted by CVPR-2023.

What's New

[04 Mar 2023]

• We support DistributedDataParallel DDP which enables data parallel training.

Installation

git clone https://github.com/XKMar/FastGait.git
cd FastGait

Example #1:

Training on CASIA-B dataset .

• STEP 1: Prepare the dataset and copy the directory into the config file DATA_ROOT (configs\benchmarks\casia-b_mixnet_gmpa.yaml).
• STEP 2: Modify the directory where the log and model files LOGS_ROOT (configs\benchmarks\casia-b_mixnet_gmpa.yaml).

Train

bash tools dist_train.sh

Test

bash tools dist_test.sh

Citation

If you find this code useful for your research, please cite our paper

@inproceedings{ma2023dynamic,
  title={Dynamic aggregated network for gait recognition},
  author={Ma, Kang and Fu, Ying and Zheng, Dezhi and Cao, Chunshui and Hu, Xuecai and Huang, Yongzhen},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={22076--22085},
  year={2023}
}