Self-Supervised Monocular Depth Estimation With Geometric Prior and Pixel-Level Sensitivity

doi:10.1109/TIV.2022.3210274

CORC > 自动化研究所 > 中国科学院自动化研究所 > 多模态人工智能系统全国重点实验室

	Self-Supervised Monocular Depth Estimation With Geometric Prior and Pixel-Level Sensitivity
	Liu, Jierui 1,3; Cao, Zhiqiang 1,3; Liu, Xilong 1,3; Wang, Shuo 1,3; Yu, Junzhi 2
刊名	IEEE TRANSACTIONS ON INTELLIGENT VEHICLES
	2023-03-01
卷号	8 期号:3 页码:2244-2256
关键词	Estimation Costs Training Sensitivity Cameras Optical flow Semantics Monocular depth estimation self-supervised learning prior feature consistency sensitivity adaptation
ISSN号	2379-8858
DOI	10.1109/TIV.2022.3210274
通讯作者	Liu, Xilong(xilong.liu@ia.ac.cn)
英文摘要	Self-supervised monocular depth estimation has gained popularity due to its convenience of training network without dense ground truth depth annotation. Specifically, the multi-frame monocular depth estimation achieves promising results in virtue of temporal information. However, existing multi-frame solutions ignore the different impacts of pixels of input frame on depth estimation and the geometric information is still insufficiently explored. In this paper, a self-supervised monocular depth estimation framework with geometric prior and pixel-level sensitivity is proposed. Geometric constraint is involved through a geometric pose estimator with prior depth predictor and optical flow predictor. Further, an alternative learning strategy is designed to improve the learning of prior depth predictor by decoupling it with the ego-motion from the geometric pose estimator. On this basis, prior feature consistency regularization is introduced into the depth encoder. By taking the dense prior cost volume based on optical flow map and ego-motion as the supervising signal for feature consistency learning, the cost volume is obtained with more reasonable feature matching. To deal with the pixel-level difference of sensitivity in input frame, a sensitivity-adaptive depth decoder is built by flexibly adding a shorter path from cost volume to the final depth prediction. In this way, the back propagation of gradient to cost volume is adaptively adjusted, and an accurate depth map is decoded. The effectiveness of the proposed method is verified on public datasets.
资助项目	National Natural Science Foundation of China[62073322] ; National Natural Science Foundation of China[61836015]
WOS研究方向	Computer Science ; Engineering ; Transportation
语种	英语
出版者	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
WOS记录号	WOS:000981348100022
资助机构	National Natural Science Foundation of China
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/53350]
专题	多模态人工智能系统全国重点实验室
通讯作者	Liu, Xilong
作者单位	1.Univ Chinese Acad Sci, Sch Artificial Intelligence, Beijing 100049, Peoples R China 2.Peking Univ, Coll Engn, Dept Adv Mfg & Robot, State Key Lab Turbulence & Complex Syst, Beijing, Peoples R China 3.Chinese Acad Sci, Inst Automat, State Key Lab Management & Control Complex Syst, Beijing 100190, Peoples R China
推荐引用方式 GB/T 7714	Liu, Jierui,Cao, Zhiqiang,Liu, Xilong,et al. Self-Supervised Monocular Depth Estimation With Geometric Prior and Pixel-Level Sensitivity[J]. IEEE TRANSACTIONS ON INTELLIGENT VEHICLES,2023,8(3):2244-2256.
APA	Liu, Jierui,Cao, Zhiqiang,Liu, Xilong,Wang, Shuo,&Yu, Junzhi.(2023).Self-Supervised Monocular Depth Estimation With Geometric Prior and Pixel-Level Sensitivity.IEEE TRANSACTIONS ON INTELLIGENT VEHICLES,8(3),2244-2256.
MLA	Liu, Jierui,et al."Self-Supervised Monocular Depth Estimation With Geometric Prior and Pixel-Level Sensitivity".IEEE TRANSACTIONS ON INTELLIGENT VEHICLES 8.3(2023):2244-2256.