notesum.ai

Published at November 21

Stereo Anything: Unifying Stereo Matching with Large-Scale Mixed Data

cs.CV

Released Date: November 21, 2024

Authors: Xianda Guo¹, Chenming Zhang², Youmin Zhang³, Dujun Nie⁴, Ruilin Wang⁴, Wenzhao Zheng⁵, Matteo Poggi³, Long Chen⁴

Aff.: ¹School of Computer Science, Wuhan University; ²Institute of Artificial Intelligence and Robotics, Xi'an Jiaotong University; ³University of Bologna; ⁴Institute of Automation, Chinese Academy of Sciences; ⁵University of California, Berkeley

Arxiv: http://arxiv.org/abs/2411.14053v1

Method	K12	K15	Midd	E3D	DR	Mean
PSMNet [6]	30.51	32.15	33.53	18.02	36.19	30.08
CFNet [36]	13.64	12.09	23.91	7.67	27.26	16.91
GwcNet [14]	23.05	25.19	29.87	14.54	35.40	25.61
COEX [1]	12.08	11.01	25.17	11.43	24.17	16.77
FADNet++ [41]	11.31	13.23	24.07	22.48	20.50	18.32
Cascade [12]	11.86	12.06	27.39	11.62	27.65	18.12
LightStereo-L	6.41	6.40	17.51	11.33	21.74	12.68
IGEV [46]	4.88	5.16	8.47	3.53	6.2	5.67
StereoBase [15]	4.85	5.35	9.76	3.12	11.84	6.98
NMRFStereo [13]	4.20	5.10	7.50	3.80	11.92	6.50
NMRFStereo [13]^∗	8.67	7.46	16.36	23.46	34.58	18.11
StereoAnything	4.29	4.31	6.96	1.84	7.64	5.01