Distribution-Guided Hierarchical Calibration Contrastive Network for Unsupervised Person Re-Identification
This repository contains the author's implementation in PyTorch for the TCSVT paper "Distribution-Guided Hierarchical Calibration Contrastive Network for Unsupervised Person Re-Identification"
pip install -r requirements.txt
##Prepare Datasets
cd examples && mkdir data
Download the person datasets Market-1501,MSMT17,PersonX,DukeMTMC-reID and the vehicle datasets VeRi-776 from aliyun. Then unzip them under the directory like
ClusterContrast/examples/data
├── market1501
│ └── Market-1501-v15.09.15
├── msmt17
│ └── MSMT17_V1
├── personx
│ └── PersonX
├── dukemtmcreid
│ └── DukeMTMC-reID
We utilize 4 GTX-3090 GPUs for training.
examples:
Market-1501:
CUDA_VISIBLE_DEVICES=0,1,2,3 python examples/dhccn.py -b 256 -a hfe -d market1501 --iters 200 --momentum 0.4 --eps 0.4 --lamb=0.5 --noisy-threshold 0.1 --num-instances 16
MSMT17:
CUDA_VISIBLE_DEVICES=0,1,2,3 python examples/dhccn.py -b 256 -a hfe -d msmt17 --iters 400 --momentum 0.3 --eps 0.7 --lamb 0.7 --noisy-threshold 0.1 --num-instances 16
DukeMTMC-reID:
CUDA_VISIBLE_DEVICES=0,1,2,3 python examples/dhccn.py -b 256 -a resnet50 -d dukemtmcreid --iters 200 --momentum 0.1 --eps 0.6 --lamb 0.5 --noisy-threshold 0.1 --num-instances 16
Personx:
CUDA_VISIBLE_DEVICES=0,1,2,3 python examples/dhccn.py -b 256 -a resnet50 -d personx --iters 200 --momentum 0.1 --eps 0.7 --lamb 0.5 --noisy-threshold 0.1 --num-instances 16
The code is based on CCL licensed under MIT.