notesum.ai

Published at October 21

GReFEL: Geometry-Aware Reliable Facial Expression Learning under Bias and Imbalanced Data Distribution

cs.SE

cs.AI

cs.HC

cs.LG

Released Date: October 21, 2024

Authors: Azmine Toushik Wasi¹, Taki Hasan Rafi², Raima Islam³, Karlo Serbetar⁴, Dong Kyu Chae

Aff.: ¹Shahjalal University of Science and Technology, Bangladesh; ²Hanyang University, South Korea; ³Harvard University, USA; ⁴University of Cambridge, United Kingdom

Arxiv: https://arxiv.org/abs/2410.15927v1

Models	Datasets
Models	AffectNet^∗†	AffWild2^∗†	RAF-DB^∗†	JAFFE^∗	FER+^∗†	FERG
SCN [34] (CVPR’20)	56.35	60.55	87.03	86.33	85.97	90.46
RAN [33] (TIP’20)	52.97	59.81	86.90	88.67	83.63	90.22
DMUE [30] (CVPR’21)	61.21	63.64	83.19	-	-	-
Tr.FER [39] (ICCV’21)	66.23	68.92	90.91	-	-	-
RUL [42] (NIPS’21)	60.65	62.37	88.98	92.33	-	92.35
Eff.Face [46] (AAAI’21)	59.89	62.21	88.36	92.33	-	92.16
F2Exp [41] (CVPR’22)	64.23	66.34	88.54	-	-	-
POSTER [47] (IC-W’22)	63.34	67.74	92.05	94.57	91.62	95.82
EAC [43] (ECCV’22)	61.11	63.54	88.02	-	87.03	-
L.OFER [14] (ICCV’23)	63.90	66.02	89.60	-	-	-
LA-Net [38] (ICCV’23)	64.54	66.76	91.56	-	91.78	-
DAN [36] (Bioinf.’23)	62.09	65.82	89.70	-	-	-
POSTER⁺⁺ [26] (’23)	63.76	69.18	92.21	96.67	92.28	96.36
GReFEL (Ours)	68.02	72.48	92.47	96.67	93.09	98.18