Panoramic depth estimation via supervised and unsupervised learning in indoor scenes

Keyang Zhou; Kailun Yang; Kaiwei Wang; Kaiwei Wang

doi:10.1364/AO.432534

Applied Optics
Vol. 60,
Issue 26,
pp. 8188-8197
(2021)
•https://doi.org/10.1364/AO.432534

Panoramic depth estimation via supervised and unsupervised learning in indoor scenes

Keyang Zhou, Kailun Yang, and Kaiwei Wang

Not Accessible

Your library or personal account may give you access

Get PDF
Email
Share
Get Citation
Copy Citation Text
Keyang Zhou, Kailun Yang, and Kaiwei Wang, "Panoramic depth estimation via supervised and unsupervised learning in indoor scenes," Appl. Opt. 60, 8188-8197 (2021)

Export Citation
- BibTex
- Endnote (RIS)
- HTML
- Plain Text
Citation alert
Save article

Check for updates

Abstract

Depth estimation, as a necessary clue to convert 2D images into the 3D space, has been applied in many machine vision areas. However, to achieve an entire surrounding 360° geometric sensing, traditional stereo matching algorithms for depth estimation are limited due to large noise, low accuracy, and strict requirements for multi-camera calibration. In this work, for a unified surrounding perception, we introduce panoramic images to obtain a larger field of view. We extend PADENet [IEEE 23rd International Conference on Intelligent Transportation Systems, (2020), pp. 1–6 [CrossRef] ], which first appeared in our previous conference work for outdoor scene understanding, to perform panoramic monocular depth estimation with a focus for indoor scenes. At the same time, we improve the training process of the neural network adapted to the characteristics of panoramic images. In addition, we fuse the traditional stereo matching algorithm with deep learning methods and further improve the accuracy of depth predictions. With a comprehensive variety of experiments, this research demonstrates the effectiveness of our schemes aiming for indoor scene perception.

Full Article | PDF Article

More Like This

Self-supervised stereo depth estimation based on bi-directional pixel-movement learning

Huachun Wang, Xinzhu Sang, Duo Chen, Peng Wang, Xiaoqian Ye, Shuai Qi, and Binbin Yan
Appl. Opt. 61(7) D7-D14 (2022)

Monocular catadioptric panoramic depth estimation via caustics-based virtual scene transition

Yu He, Lingxue Wang, Yi Cai, and Wei Xue
J. Opt. Soc. Am. A 33(9) 1872-1879 (2016)

Binocular stereo matching of real scenes based on a convolutional neural network and computer graphics

Liaoyu Kou, Kai Yang, Lin Luo, Yu Zhang, Jinlong Li, Yong Wang, and Liming Xie
Opt. Express 29(17) 26876-26893 (2021)

Previous Article Next Article

Data Availability

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.

Cited By

You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Figures (8)

You do not have subscription access to this journal. Figure files are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Tables (2)

You do not have subscription access to this journal. Article tables are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Equations (14)

You do not have subscription access to this journal. Equations are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Methods	Abs. Rel.	Sq. Rel.	RMSE	RMSE log	Acc: $δ < 1.25$	Acc: $δ < {1.25}^{2}$	Acc: $δ < {1.25}^{3}$
Nikolaos et al. [34]	0.138	0.091	0.473	0.184	82.4%	95.9%	98.9%
Supervised learning	0.061	0.036	0.366	0.132	91.3%	98.5%	99.5%
Unsupervised learning	0.070	0.043	0.402	0.177	88.7%	97.4%	99.0%
Fused learning	0.050	0.027	0.326	0.121	94.0%	98.9%	99.7%

Methods	Abs. Rel.	Sq. Rel.	RMSE	RMSE log	Acc: $δ < 1.25$	Acc: $δ < {1.25}^{2}$	Acc: $δ < {1.25}^{3}$
Supervised Learning with SGM
w/o SGM	0.167	0.190	0.826	0.294	38.6%	72.0%	97.4%
w SGM	0.167	0.239	0.800	0.252	50.3%	89.4%	96.6%
Unsupervised Learning with SGM
w/o SGM	0.187	0.223	0.936	0.348	31.8%	62.2%	94.0%
w SGM	0.183	0.240	0.841	0.298	42.8%	79.7%	95.5%
Fused Learning with SGM
w/o SGM	0.162	0.178	0.812	0.287	39.5%	73.7%	97.9%
w SGM	0.163	0.223	0.771	0.246	51.3%	90.1%	96.9%

Methods	Abs. Rel.	Sq. Rel.	RMSE	RMSE log	Acc: $δ < 1.25$	Acc: $δ < {1.25}^{2}$	Acc: $δ < {1.25}^{3}$
Nikolaos et al. [34]	0.138	0.091	0.473	0.184	82.4%	95.9%	98.9%
Supervised learning	0.061	0.036	0.366	0.132	91.3%	98.5%	99.5%
Unsupervised learning	0.070	0.043	0.402	0.177	88.7%	97.4%	99.0%
Fused learning	0.050	0.027	0.326	0.121	94.0%	98.9%	99.7%

Methods	Abs. Rel.	Sq. Rel.	RMSE	RMSE log	Acc: $δ < 1.25$	Acc: $δ < {1.25}^{2}$	Acc: $δ < {1.25}^{3}$
Supervised Learning with SGM
w/o SGM	0.167	0.190	0.826	0.294	38.6%	72.0%	97.4%
w SGM	0.167	0.239	0.800	0.252	50.3%	89.4%	96.6%
Unsupervised Learning with SGM
w/o SGM	0.187	0.223	0.936	0.348	31.8%	62.2%	94.0%
w SGM	0.183	0.240	0.841	0.298	42.8%	79.7%	95.5%
Fused Learning with SGM
w/o SGM	0.162	0.178	0.812	0.287	39.5%	73.7%	97.9%
w SGM	0.163	0.223	0.771	0.246	51.3%	90.1%	96.9%

Abstract

Data Availability

Cited By

Figures (8)

Tables (2)

Equations (14)

Applied Optics