Home

Awesome

This is the official repository of the paper:

ElasticFace: Elastic Margin Loss for Deep Face Recognition

Paper on arxiv: arxiv

*** Accepted CVPR workshops 2022 ***

evaluation

ModelLog filePretrained modelcheckpoint
ElasticFace-Arclog filepretrained-mode295672backbone.pth
ElasticFace-Coslog filepretrained-mode295672backbone.pth
ElasticFace-Arc+log filepretrained-mode295672backbone.pth
ElasticFace-Cos+log filepretrained-mode295672backbone.pth

Evaluation result: See: Paper with code

Face recognition model training

Model training: In the paper, we employ MS1MV2 as the training dataset which can be downloaded from InsightFace (MS1M-ArcFace in DataZoo) Download MS1MV2 dataset from insightface on strictly follow the licence distribution

Unzip the dataset and place it in the data folder Set the config.output and config.loss in the config/config.py

All code has been trained and tested using Pytorch 1.7.1

Face recognition evaluation

evaluation on LFW, AgeDb-30, CPLFW, CALFW and CFP-FP:
  1. download the data from their offical webpages.
  2. alternative: The evaluation datasets are available in the training dataset package as bin file
  3. set the config.rec to dataset folder e.g. data/faces_emore
  4. set the config.val_targets for list of the evaluation dataset
  5. download the pretrained model from link the previous table
  6. set the config.output to path to pretrained model weights
  7. run eval/evaluation.py
  8. the output is test.log contains the evaluation results over all epochs

To-do

If you use any of the code provided in this repository, please cite the following paper:

Citation

@InProceedings{Boutros_2022_CVPR,
    author    = {Boutros, Fadi and Damer, Naser and Kirchbuchner, Florian and Kuijper, Arjan},
    title     = {ElasticFace: Elastic Margin Loss for Deep Face Recognition},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
    month     = {June},
    year      = {2022},
    pages     = {1578-1587}
}


License

This project is licensed under the terms of the Attribution-NonCommercial-ShareAlike 4.0 
International (CC BY-NC-SA 4.0) license. 
Copyright (c) 2021 Fraunhofer Institute for Computer Graphics Research IGD Darmstadt