katerina-merkulova / simple-reid

A Simple Codebase for Image-based Person Re-identification

Requirements: Python 3.6, Pytorch 1.6.0, yacs

Supported losses

Classification Losses

• ArcFace Loss

Pairwise Losses

• Triplet Loss

Supported models

• ResNet-50

Get Started

• Replace _C.DATA.ROOT and _C.OUTPUT in configs/default.py with your own data path and output path, respectively.
• Run train.sh

Citation

If you use our code in your research or wish to refer to the baseline results, please use the following BibTeX entry.

@InProceedings{CVPR2019IANet
author = {Hou, Ruibing and Ma, Bingpeng and Chang, Hong and Gu, Xinqian and Shan, Shiguang and Chen, Xilin},
title = {Interaction-And-Aggregation Network for Person Re-Identification},
booktitle = {CVPR},
year = {2019}
}