Distributed Arcface/Adaface Training in Pytorch

Modified old version of PartialFC to work with AdaFace for both normal and distributed training.

• Killing Two Birds With One Stone: Efficient and Robust Training of Face Recognition CNNs by Partial FC (CVPR-2022)
• AdaFace: Quality Adaptive Margin for Face Recognition (CVPR-2022)

Benchmark

1. CASIA-Webface (10k ids, 0.5M images)

Current have opposite view between feature norm & image quality than described in paper

• Accuracy

Model	Backbone	Sample Rate	LFW	CFP-FP	AGEDB-30	LFW Blur	CFP-FP Blur	AGEDB-30 Blur	Average
Arcface	IR50	1.0	0.9920	0.9601	0.9365	0.9323	0.8517	0.8357	0.9181
Adaface	IR50	1.0	0.9923	0.9587	0.9405	0.9563	0.8667	0.8632	0.9296
Arcface	IR50	0.3	0.9923	0.9596	0.9390	0.9323	0.8480	0.8325	0.9173
Adaface	IR50	0.3	0.9915	0.9567	0.9362	0.9532	0.8570	0.8548	0.9249

• Features Norm

Model	Backbone	Sample Rate	LFW	CFP-FP	AGEDB-30	LFW Blur	CFP-FP Blur	AGEDB-30 Blur
Arcface	IR50	1.0	12.72	12.97	13.23	12.56	13.16	12.86
Adaface	IR50	1.0	5.34	7.9	5.91	10.36	42.41	10.99
Arcface	IR50	0.3	14.3	14.1	14.54	14	14.68	14.29
Adaface	IR50	0.3	6.09	10.23	6.74	10.7	47.28	11.7

How to Training

To train a model, run train.py with the path to the configs:

1. Single node, 8 GPUs:

python -m torch.distributed.launch --nproc_per_node=8 --nnodes=1 --node_rank=0 --master_addr="127.0.0.1" --master_port=1234 train.py configs/ms1mv3_r50-adaface.py

2. Multiple nodes, each node 8 GPUs (not tested with AdaFace):

Node 0:

python -m torch.distributed.launch --nproc_per_node=8 --nnodes=2 --node_rank=0 --master_addr="ip1" --master_port=1234 train.py train.py configs/ms1mv3_r50-adaface.py

Node 1:

python -m torch.distributed.launch --nproc_per_node=8 --nnodes=2 --node_rank=1 --master_addr="ip1" --master_port=1234 train.py train.py configs/ms1mv3_r50-adaface.py

To do

• Add ViT models
• Report comparision between Adaface, Arcface & Cosface (currently in training, dataset 5M ids and 100M images)
• Result for common large scale face recognition dataset (MS1MV2, MS1MV3, Glint360k, WebFace)

Reference

• Insightface
• Adaface