notesum.ai

Published at November 29

cs.CV

cs.LG

Released Date: November 29, 2024

Authors: Qirui Wu¹, Denys Iliash¹, Daniel Ritchie², Manolis Savva¹, Angel X. Chang³

Aff.: ¹Simon Fraser University; ²Brown University; ³Simon Fraser University, Alberta Machine Intelligence Institute (Amii)

	Ground Truth Depth						Estimated Depth
Zero-shot pose estimation	Scene-aware Alignment $\uparrow$				Collision $\downarrow$	Relation $\uparrow$	Scene-aware Alignment $\uparrow$				Collision $\downarrow$	Relation $\uparrow$
Zero-shot pose estimation	rAcc	tAcc	sAcc	Acc	Collision $\downarrow$	Relation $\uparrow$	rAcc	tAcc	sAcc	Acc	Collision $\downarrow$	Relation $\uparrow$
Best-matching multiview	0.34	0.93	0.52	0.19	11.43	0.58	0.34	0.84	0.44	0.15	12.29	0.25
ZSP [30] w/ DinoV2	0.37	0.93	0.57	0.23	10.48	0.59	0.42	0.85	0.55	0.26	8.29	0.25
ZSP [30] w/ ft DinoV2	0.45	0.94	0.66	0.34	7.90	0.61	0.41	0.85	0.54	0.24	8.46	0.24
GigaPose [63]	0.36	0.95	0.71	0.27	7.91	0.61	0.36	0.84	0.64	0.22	7.44	0.22
Ours	0.45	0.95	0.71	0.33	7.26	0.61	0.41	0.84	0.64	0.25	7.37	0.25