notesum.ai

Published at November 22

Adaptive Hyper-Graph Convolution Network for Skeleton-based Human Action Recognition with Virtual Connections

cs.CV

Released Date: November 22, 2024

Authors: Youwei Zhou¹, Tianyang Xu¹, Cong Wu¹, Xiaojun Wu¹, Josef Kittler²

Aff.: ¹School of Artificial Intelligence and Computer Science, Jiangnan University; ²School of Computer Science and Electronic Engineering, and CVSSP, University of Surrey

Arxiv: http://arxiv.org/abs/2411.14796v1

Methods	Publication	Modalities	GFLOPs	Params (M)	NTU-RGB+D 60		NTU-RGB+D 120		NW-UCLA (%)
Methods	Publication	Modalities	GFLOPs	Params (M)	X-Sub (%)	X-View (%)	X-Sub (%)	X-Set (%)	NW-UCLA (%)
ST-GCN [18]	AAAI 2018	J+B	-	-	81.5	88.3	70.7	73.2	-
2s-AGCN [20]	CVPR 2019	J+B	-	-	88.5	95.1	82.5	84.2	-
DGNN [32]	CVPR 2019	J+B+JM+BM	-	-	89.9	96.1	-	-	-
SGN [30]	CVPR 2020	J+B+JM+BM	-	-	89.0	94.5	79.2	81.5	92.5
DC-GCN+ADG [33]	ECCV 2020	J+B+JM+BM	1.83	4.9	90.8	96.6	86.5	88.1	95.3
DDGCN [31]	ECCV 2020	J+B+JM+BM	-	-	91.1	97.1	-	-	-
MS-G3D [42]	CVPR 2020	J+B+JM+BM	5.22	2.8	91.5	96.2	86.9	88.4	-
MST-GCN [34]	AAAI 2021	J+B+JM+BM	-	12.0	91.5	96.6	87.5	88.8	-
CTR-GCN [13]	ICCV 2021	J+B+JM+BM	1.97	1.5	92.4	96.4	88.9	90.6	96.5
EfficientGCN-B4 [43]	TPAMI 2022	J+B+JM+BM	15.2	2.0	91.7	95.7	88.3	89.1	-
InfoGCN* [12]	CVPR 2022	J+B+J’+B’	1.84	1.6	92.7	96.9	89.4	90.7	96.6
FR Head [48]	CVPR 2023	J+B+JM+BM	-	2.0	92.8	96.8	89.5	90.9	96.8
HD-GCN* [19]	ICCV 2023	J+B+J’+B’	1.77	1.7	93.0	97.0	89.8	91.2	96.9
DS-GCN [14]	AAAI 2024	J+B+JM+BM	-	-	93.1	97.5	89.2	90.3	-
Hyper-GCN (w/o ensemble)	-	J	1.62	1.0	91.1	95.3	87.0	88.4	94.8
Hyper-GCN (2-ensemble)	-	J+B	1.62	1.0	92.9	96.3	89.8	90.9	96.2
Hyper-GCN (4-ensemble)	-	J+B+JM+BM	1.62	1.0	93.1	96.7	90.2	91.4	96.6