notesum.ai

Published at December 9

Adaptive Graph Learning from Spatial Information for Surgical Workflow Anticipation

cs.CV

cs.RO

Released Date: December 9, 2024

Authors: Francis Xiatian Zhang¹, Jingjing Deng¹, Robert Lieck¹, Hubert P. H. Shum¹

Aff.: ¹Durham University, UK

Arxiv: http://arxiv.org/pdf/2412.06454v1

Refer to caption

$wMAE$ Comparison	Instrument Anticipation				Phase Anticipation
$wMAE$ Comparison	2 min	3 min	5 min	$\text{Mean}_{2,3,5}$	2 min	3 min	5 min	$\text{Mean}_{2,3,5}$
TimeLSTM [3]	0.51	0.76	1.32	0.86	0.47	0.64	1.07	0.72
RSDNet [4]	0.48	0.73	1.26	0.82	0.43	0.63	1.10	0.72
TempAgg [43]	0.65	0.92	1.47	1.01	0.38	0.50	1.18	0.68
B-CNN-LSTM[1]	0.43	0.66	1.09	0.73	0.39	0.59	0.85	0.61
IIA-Net [6]	0.38	0.58	0.92	0.63	0.36	0.49	0.68	0.51
GCN-MSTCN [11]	0.48	0.72	1.21	0.80	0.45	0.67	1.06	0.73
Ours	0.39	0.57	1.03	0.66	0.35	0.47	0.80	0.54
w/o AG	0.45	0.70	1.33	0.83	0.45	0.66	1.16	0.76
w/o MHO	0.43	0.61	1.00	0.68	0.37	0.51	0.80	0.56