Awesome

META

[ECCV 2022] Mimic Embedding via Adaptive Aggregation: Learning Generalizable Person Re-identification [paper]

Update

2022-07-14: Update Code.

2022-08-03: Update Paper.

Datasets

• Requirements: Market1501, CUHK03, CUHK-SYSU, MSMT17_v2

Please put all the datasets in one directory, and modify DATASETS in projects/META/configs/Base-cnn.yml

Installation

• Please check fast-reid for fast-reid installation
• Please check Apex for Apex installation
• Compile with cython to accelerate evalution:

bash cd fastreid/evaluation/rank_cylib; make all

Train

If you want to train with 1-GPU, run:

python projects/META/train_net.py --config-file projects/META/configs/r50.yml MODEL.DEVICE "cuda:0"

If you want to train with 4-GPU, run:

CUDA_VISIBLE_DEVICES=0,1,2,3 python projects/META/train_net.py --config-file projects/META/configs/r50.yml --num-gpus 4

You can get the results in our paper by training with 4-GPU, please modify SOLVER.IMS_PER_BATCH in projects/META/configs/Base-cnn.yml (64 for 1-GPU and 256 for 4-GPU)

Evaluation

To evaluate a model's performance, use:

python projects/META/train_net.py --config-file projects/META/configs/r50.yml --eval-only MODEL.WEIGHTS /path/to/checkpoint_file MODEL.DEVICE "cuda:0"

Contacts

If you have any question about the project, please feel free to contact me.

E-mail: boqiang.xu@cripac.ia.ac.cn

ACKNOWLEDGEMENTS

The code was developed based on the ’fast-reid’ toolbox https://github.com/JDAI-CV/fast-reid.