TASK	DATASET	MODEL	METRIC NAME	METRIC VALUE	GLOBAL RANK
Lightweight Face Recognition	AgeDB-30	EdgeFace - XS (g=0.6)	MParams	1.77	# 1
Lightweight Face Recognition	AgeDB-30	EdgeFace - XS (g=0.6)	MFLOPs	154	# 1
Lightweight Face Recognition	AgeDB-30	EdgeFace - XS (g=0.6)	Accuracy	0.96	# 4
Lightweight Face Recognition	AgeDB-30	EdgeFace - S (g=0.5)	MParams	3.65	# 3
Lightweight Face Recognition	AgeDB-30	EdgeFace - S (g=0.5)	MFLOPs	306.11	# 2
Lightweight Face Recognition	AgeDB-30	EdgeFace - S (g=0.5)	Accuracy	0.9693	# 1
Lightweight Face Recognition	CALFW	EdgeFace - S (g=0.5)	Accuracy	0.9571	# 1
Lightweight Face Recognition	CALFW	EdgeFace - S (g=0.5)	MFLOPs	306.11	# 2
Lightweight Face Recognition	CALFW	EdgeFace - S (g=0.5)	MParams	3.65	# 2
Lightweight Face Recognition	CALFW	EdgeFace - XS (g=0.6)	Accuracy	0.9528	# 3
Lightweight Face Recognition	CALFW	EdgeFace - XS (g=0.6)	MFLOPs	154	# 1
Lightweight Face Recognition	CALFW	EdgeFace - XS (g=0.6)	MParams	1.77	# 1
Lightweight Face Recognition	CFP-FP	EdgeFace - S (g=0.5)	Accuracy	0.9581	# 1
Lightweight Face Recognition	CFP-FP	EdgeFace - S (g=0.5)	MFLOPs	306.11	# 2
Lightweight Face Recognition	CFP-FP	EdgeFace - S (g=0.5)	MParams	3.65	# 3
Face Recognition	CFP-FP	EdgeFace - S (g=0.5)	Accuracy	0.9581	# 5
Lightweight Face Recognition	CFP-FP	EdgeFace - XS (g=0.6)	Accuracy	0.9437	# 2
Lightweight Face Recognition	CFP-FP	EdgeFace - XS (g=0.6)	MFLOPs	154	# 1
Lightweight Face Recognition	CFP-FP	EdgeFace - XS (g=0.6)	MParams	1.77	# 1
Lightweight Face Recognition	CPLFW	EdgeFace - S (g=0.5)	Accuracy	0.9256	# 1
Lightweight Face Recognition	CPLFW	EdgeFace - S (g=0.5)	MFLOPs	306.11	# 2
Lightweight Face Recognition	CPLFW	EdgeFace - S (g=0.5)	MParams	3.65	# 2
Lightweight Face Recognition	CPLFW	EdgeFace - XS (g=0.6)	Accuracy	0.9182	# 2
Lightweight Face Recognition	CPLFW	EdgeFace - XS (g=0.6)	MFLOPs	154	# 1
Lightweight Face Recognition	CPLFW	EdgeFace - XS (g=0.6)	MParams	1.77	# 1
Lightweight Face Recognition	IJB-B	EdgeFace - S (g=0.5)	TAR @ FAR=0.01	0.9358	# 1
Lightweight Face Recognition	IJB-B	EdgeFace - S (g=0.5)	MFLOPs	306.11	# 2
Lightweight Face Recognition	IJB-B	EdgeFace - S (g=0.5)	MParams	3.65	# 2
Lightweight Face Recognition	IJB-B	EdgeFace - XS (g=0.6)	TAR @ FAR=0.01	0.9267	# 2
Lightweight Face Recognition	IJB-B	EdgeFace - XS (g=0.6)	MFLOPs	154	# 1
Lightweight Face Recognition	IJB-B	EdgeFace - XS (g=0.6)	MParams	1.77	# 1
Lightweight Face Recognition	IJB-C	EdgeFace - S (g=0.5)	TAR @ FAR=0.01	0.9563	# 1
Lightweight Face Recognition	IJB-C	EdgeFace - S (g=0.5)	MFLOPs	306.11	# 2
Lightweight Face Recognition	IJB-C	EdgeFace - S (g=0.5)	MParams	3.65	# 2
Lightweight Face Recognition	IJB-C	EdgeFace - XS (g=0.6)	TAR @ FAR=0.01	0.9485	# 2
Lightweight Face Recognition	IJB-C	EdgeFace - XS (g=0.6)	MFLOPs	154	# 1
Lightweight Face Recognition	IJB-C	EdgeFace - XS (g=0.6)	MParams	1.77	# 1
Face Recognition	LFW	EdgeFace - S (g=0.5)	Accuracy	0.9978	# 7
Lightweight Face Recognition	LFW	EdgeFace - XS (g=0.6)	Accuracy	0.9973	# 2
Lightweight Face Recognition	LFW	EdgeFace - XS (g=0.6)	MFLOPs	154	# 1
Lightweight Face Recognition	LFW	EdgeFace - XS (g=0.6)	MParams	1.77	# 1
Lightweight Face Recognition	LFW	EdgeFace - S (g=0.5)	Accuracy	0.9978	# 1
Lightweight Face Recognition	LFW	EdgeFace - S (g=0.5)	MFLOPs	306.11	# 2
Lightweight Face Recognition	LFW	EdgeFace - S (g=0.5)	MParams	3.65	# 4
Face Recognition	LFW	EdgeFace - XS (g=0.6)	Accuracy	0.9973	# 8

Badge	Markdown
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/edgeface-efficient-face-recognition-model-for/lightweight-face-recognition-on-agedb-30)](https://paperswithcode.com/sota/lightweight-face-recognition-on-agedb-30?p=edgeface-efficient-face-recognition-model-for)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/edgeface-efficient-face-recognition-model-for/lightweight-face-recognition-on-calfw)](https://paperswithcode.com/sota/lightweight-face-recognition-on-calfw?p=edgeface-efficient-face-recognition-model-for)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/edgeface-efficient-face-recognition-model-for/lightweight-face-recognition-on-cfp-fp)](https://paperswithcode.com/sota/lightweight-face-recognition-on-cfp-fp?p=edgeface-efficient-face-recognition-model-for)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/edgeface-efficient-face-recognition-model-for/lightweight-face-recognition-on-cplfw)](https://paperswithcode.com/sota/lightweight-face-recognition-on-cplfw?p=edgeface-efficient-face-recognition-model-for)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/edgeface-efficient-face-recognition-model-for/lightweight-face-recognition-on-ijb-b)](https://paperswithcode.com/sota/lightweight-face-recognition-on-ijb-b?p=edgeface-efficient-face-recognition-model-for)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/edgeface-efficient-face-recognition-model-for/lightweight-face-recognition-on-ijb-c)](https://paperswithcode.com/sota/lightweight-face-recognition-on-ijb-c?p=edgeface-efficient-face-recognition-model-for)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/edgeface-efficient-face-recognition-model-for/lightweight-face-recognition-on-lfw)](https://paperswithcode.com/sota/lightweight-face-recognition-on-lfw?p=edgeface-efficient-face-recognition-model-for)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/edgeface-efficient-face-recognition-model-for/face-recognition-on-cfp-fp)](https://paperswithcode.com/sota/face-recognition-on-cfp-fp?p=edgeface-efficient-face-recognition-model-for)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/edgeface-efficient-face-recognition-model-for/face-recognition-on-lfw)](https://paperswithcode.com/sota/face-recognition-on-lfw?p=edgeface-efficient-face-recognition-model-for)`

EdgeFace: Efficient Face Recognition Model for Edge Devices

4 Jul 2023 · Anjith George, Christophe Ecabert, Hatef Otroshi Shahreza, Ketan Kotwal, Sebastien Marcel ·

In this paper, we present EdgeFace, a lightweight and efficient face recognition network inspired by the hybrid architecture of EdgeNeXt. By effectively combining the strengths of both CNN and Transformer models, and a low rank linear layer, EdgeFace achieves excellent face recognition performance optimized for edge devices. The proposed EdgeFace network not only maintains low computational costs and compact storage, but also achieves high face recognition accuracy, making it suitable for deployment on edge devices. Extensive experiments on challenging benchmark face datasets demonstrate the effectiveness and efficiency of EdgeFace in comparison to state-of-the-art lightweight models and deep face recognition models. Our EdgeFace model with 1.77M parameters achieves state of the art results on LFW (99.73%), IJB-B (92.67%), and IJB-C (94.85%), outperforming other efficient models with larger computational complexities. The code to replicate the experiments will be made available publicly.

PDF Abstract

Code

Add Remove Mark official

otroshi/edgeface official

anjith2006/bob.paper.tbiom2023_edge…

bob/bob.paper.tbiom2023_edgeface

Tasks

Add Remove

Face Recognition

Lightweight Face Recognition

Datasets

LFW

CASIA-WebFace

IJB-C

IJB-B CPLFW CALFW

Results from the Paper

Edit

Ranked #1 on Lightweight Face Recognition on CFP-FP

Get a GitHub badge

Task	Dataset	Model	Metric Name	Metric Value	Global Rank	Benchmark
Lightweight Face Recognition	AgeDB-30	EdgeFace - XS (g=0.6)	MParams	1.77	# 1	Compare
			MFLOPs	154	# 1	Compare
			Accuracy	0.96	# 4	Compare
Lightweight Face Recognition	AgeDB-30	EdgeFace - S (g=0.5)	MParams	3.65	# 3	Compare
			MFLOPs	306.11	# 2	Compare
			Accuracy	0.9693	# 1	Compare
Lightweight Face Recognition	CALFW	EdgeFace - S (g=0.5)	Accuracy	0.9571	# 1	Compare
			MFLOPs	306.11	# 2	Compare
			MParams	3.65	# 2	Compare
Lightweight Face Recognition	CALFW	EdgeFace - XS (g=0.6)	Accuracy	0.9528	# 3	Compare
			MFLOPs	154	# 1	Compare
			MParams	1.77	# 1	Compare
Lightweight Face Recognition	CFP-FP	EdgeFace - S (g=0.5)	Accuracy	0.9581	# 1	Compare
			MFLOPs	306.11	# 2	Compare
			MParams	3.65	# 3	Compare
Face Recognition	CFP-FP	EdgeFace - S (g=0.5)	Accuracy	0.9581	# 5	Compare
Lightweight Face Recognition	CFP-FP	EdgeFace - XS (g=0.6)	Accuracy	0.9437	# 2	Compare
			MFLOPs	154	# 1	Compare
			MParams	1.77	# 1	Compare
Lightweight Face Recognition	CPLFW	EdgeFace - S (g=0.5)	Accuracy	0.9256	# 1	Compare
			MFLOPs	306.11	# 2	Compare
			MParams	3.65	# 2	Compare
Lightweight Face Recognition	CPLFW	EdgeFace - XS (g=0.6)	Accuracy	0.9182	# 2	Compare
			MFLOPs	154	# 1	Compare
			MParams	1.77	# 1	Compare
Lightweight Face Recognition	IJB-B	EdgeFace - S (g=0.5)	TAR @ FAR=0.01	0.9358	# 1	Compare
			MFLOPs	306.11	# 2	Compare
			MParams	3.65	# 2	Compare
Lightweight Face Recognition	IJB-B	EdgeFace - XS (g=0.6)	TAR @ FAR=0.01	0.9267	# 2	Compare
			MFLOPs	154	# 1	Compare
			MParams	1.77	# 1	Compare
Lightweight Face Recognition	IJB-C	EdgeFace - S (g=0.5)	TAR @ FAR=0.01	0.9563	# 1	Compare
			MFLOPs	306.11	# 2	Compare
			MParams	3.65	# 2	Compare
Lightweight Face Recognition	IJB-C	EdgeFace - XS (g=0.6)	TAR @ FAR=0.01	0.9485	# 2	Compare
			MFLOPs	154	# 1	Compare
			MParams	1.77	# 1	Compare
Face Recognition	LFW	EdgeFace - S (g=0.5)	Accuracy	0.9978	# 7	Compare
Lightweight Face Recognition	LFW	EdgeFace - XS (g=0.6)	Accuracy	0.9973	# 2	Compare
			MFLOPs	154	# 1	Compare
			MParams	1.77	# 1	Compare
Lightweight Face Recognition	LFW	EdgeFace - S (g=0.5)	Accuracy	0.9978	# 1	Compare
			MFLOPs	306.11	# 2	Compare
			MParams	3.65	# 4	Compare
Face Recognition	LFW	EdgeFace - XS (g=0.6)	Accuracy	0.9973	# 8	Compare

Methods

Add Remove

Absolute Position Encodings • Adam • BPE • Dense Connections • Dropout • Label Smoothing • Layer Normalization • Linear Layer • Multi-Head Attention • Position-Wise Feed-Forward Layer • Residual Connection • Scaled Dot-Product Attention • Softmax • Transformer

Edit Social Preview

EdgeFace: Efficient Face Recognition Model for Edge Devices

Code Edit Add Remove Mark official

Tasks Edit Add Remove

Datasets Edit

Results from the Paper Edit

Methods Edit Add Remove

Code

Add Remove Mark official

Tasks

Add Remove

Datasets

Results from the Paper

Edit

Methods

Add Remove