Pytorch implementation of CVPR2020 GroupFace , Any issue or star will be appreciated.
This repository give you a easy way to quick look through the all training process of Group Face
paper is avaliable at https://arxiv.org/abs/2005.10497
some code is ported from https://github.com/SeungyounShin/GroupFace
thanks for contributing
python train.py to launch training after 5 epochs u will found the demo set top1 acc increase obivously
- dataloader , sample ims,
- implement loss
- main process of training a demo set
- image clustering
- implement partial FC
- make backbone resNet50 and use Frelu as activation
- learning from scratch of MS1M 85K ids
- learning DGN
- train at MSCeleb[cleaned]
- eval at MegaFace LFW YTF CFP-TP AgeDB IJB-B IJB-C
- hard or soft GroupFace probability study
- group ablation study
- feature aggregation or concatenate study
- group aware similarity study