lhf12278 / DGPS

Breaking the Positive Sample Barrier in Person Re-Identification: Towards Domain Generalization without Paired Samples

Breaking the Positive Sample Barrier in Person Re-Identification: Towards Domain Generalization without Paired Samples

Pipeline

Environment

pip install -r requirement.txt

• requirement:

Python 3.9.0
Pytorch 1.10.0 & torchvision 0.11.0

Dataset Preparation

1. Download the datasets(Market-1501, MSMT17)and then unzip them to your_dataset_dir.
2. Split Market-1501 and MSMT to Market-SCT and MSMT-SCT according to CCFP.
3. Make new directories in data and organize them as follows:

+-- data
|   +-- market1501
|       +-- boudning_box_train
|       +-- query
|       +-- boudning_box_test
|   +-- market1501_sct
|       +-- boudning_box_train_sct
|       +-- query
|       +-- boudning_box_test
|   +-- MSMT17
|       +-- train_sct
|       +-- test
|       +-- list_train.txt
|       +-- MSMT_mixSCT.txt

Train and test

train

CUDA_VISIBLE_DEVICES=0 python train.py --config-file configs/Market_SCT/vit_transreid.yml

test

CUDA_VISIBLE_DEVICES=0 python test.py --config-file configs/Market_SCT/vit_transreid.yml

Contact

If you have any questions, please feel free to contact me([e-mail]).