This is the official PyTorch implementation of the ICCV 2023 paper.
UniFace: Unified Cross-Entropy Loss for Deep Face Recognition.pdf
-
Prepare dataset
Download CASIA-Webface preprocessed by insightface.
unzip faces_webface_112x112.zip
-
Train model
Modify the 'data_path' in train.py (Line 56)
Select and uncomment the 'loss' in backbone.py (Line 67)
python train.py
-
Test model
python pytorch2onnx.py zip model.zip model.onnx
Upload model.zip to MFR Ongoing and then wait for the results.
We provide a pre-trained model (ResNet-50) on Google Drive for easy and direct development. This model is trained on CASIA-WebFace and achieved 48.42% on MR-All and 99.56% on LFW.
If you find UniFace useful in your research, please consider to cite:
@InProceedings{Zhou_2023_ICCV,
author = {Zhou, Jiancan and Jia, Xi and Li, Qiufu and Shen, Linlin and Duan, Jinming},
title = {UniFace: Unified Cross-Entropy Loss for Deep Face Recognition},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
month = {October},
year = {2023},
pages = {20730-20739}
}