Awesome

Online Pseudo Label Generation by Hierarchical Cluster Dynamics for Adaptive Person Re-identification

This repository provides testing code and models of #2775.

Requirements

Installation

python setup.py install

Prepare Datasets

Download the person datasets DukeMTMC-reID, Market-1501, MSMT17. Then unzip them under the directory like

HCD/data
├── dukemtmc
│   └── DukeMTMC-reID
├── market1501
│   └── Market-1501-v15.09.15
└── msmt17
    └── MSMT17_V1

Evaluation

Unsupervised Domain Adaptation

To evaluate the model on the target-domain dataset, run:

CUDA_VISIBLE_DEVICES=0 python test.py --dsbn -d $DATASET --resume $PATH_MODEL

Example: DukeMTMC-reID -> Market-1501

CUDA_VISIBLE_DEVICES=0 python test.py --dsbn -d market1501 --resume uda_duke2market.pth.tar

Unsupervised Learning

To evaluate the model, run:

CUDA_VISIBLE_DEVICES=0 python test.py -d $DATASET --resume $PATH

Example: DukeMTMC-reID

CUDA_VISIBLE_DEVICES=0 python test.py -d dukemtmc --resume usl_duke.pth.tar

Trained Models

You can download models in the paper from Google Drive.