VSLA-CLIP++: A Large-Scale Benchmark and Part-Level Feature Alignment for Cross-Platform Video Person ReID

Installation

conda create -n vslaclip python=3.8
conda activate vslaclip
conda install pytorch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0 cudatoolkit=10.2 -c pytorch
pip install yacs
pip install timm
pip install scikit-image
pip install tqdm
pip install ftfy
pip install regex

Training

For example, if you want to run for the ls-vid, you need to modify the config file to

DATASETS:
   NAMES: ('lsvid')
   ROOT_DIR: ('your_dataset_dir')
OUTPUT_DIR: 'your_output_dir'

Then, if you want to use weight of VIFI-CLIP to initialize model, you need to down the weight form link and modify config file as:

MODEL:
  VIFI_WEIGHT : 'your_dataset_dir/vifi_weight.pth'
  USE_VIFI_WEIGHT : True

if you want to run VSLA-CLIP++:

CUDA_VISIBLE_DEVICES=0 python train_vsla_part.py --config_file configs/vsla++/vsla++_g2av2.yml

Evaluation

For example, if you want to test VSLA-CLIP++ for LS-VID

CUDA_VISIBLE_DEVICES=0 python test_vsla_part.py --config_file 'your_config_file' TEST.WEIGHT 'your_trained_checkpoints_path/ViT-B-16_120.pth'

Citation

@inproceedings{zhang2024cross,
  title={Cross-platform video person reid: A new benchmark dataset and adaptation approach},
  author={Zhang, Shizhou and Luo, Wenlong and Cheng, De and Yang, Qingchun and Ran, Lingyan and Xing, Yinghui and Zhang, Yanning},
  booktitle={European Conference on Computer Vision},
  pages={270--287},
  year={2024},
  organization={Springer}
}

Acknowledgement

Codebase from CLIP-ReID, TransReID, CLIP, and CoOp.