Self-Supervised Monocular Depth Estimation With Extensive Pretraining

Although depth estimation is a key technology for three-dimensional sensing applications involving motion, active sensors such as LiDAR and depth cameras tend to be expensive and bulky. Here, we explore the potential of monocular depth estimation (MDE) based on a self-supervised approach. MDE is a p...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autor principal:	Hyukdoo Choi
Formato:	article
Lenguaje:	EN
Publicado:	IEEE 2021
Materias:	Monocular depth estimation depth prediction convolutional neural networks self-supervised learning unsupervised learning Electrical engineering. Electronics. Nuclear engineering TK1-9971
Acceso en línea:	https://doaj.org/article/05fc891f131343b9b113a4ba39a130ca
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:05fc891f131343b9b113a4ba39a130ca
record_format	dspace
spelling	oai:doaj.org-article:05fc891f131343b9b113a4ba39a130ca2021-12-02T00:00:17ZSelf-Supervised Monocular Depth Estimation With Extensive Pretraining2169-353610.1109/ACCESS.2021.3129628https://doaj.org/article/05fc891f131343b9b113a4ba39a130ca2021-01-01T00:00:00Zhttps://ieeexplore.ieee.org/document/9622207/https://doaj.org/toc/2169-3536Although depth estimation is a key technology for three-dimensional sensing applications involving motion, active sensors such as LiDAR and depth cameras tend to be expensive and bulky. Here, we explore the potential of monocular depth estimation (MDE) based on a self-supervised approach. MDE is a promising technology, but supervised learning suffers from a need for accurate ground-truth depth data. Recent studies have enabled self-supervised training on an MDE model with only monocular image sequences and image-reconstruction errors. We pretrained networks using multiple datasets, including monocular and stereo image sequences. The main challenges posed by the self-supervised MDE model were occlusions and dynamic objects. We proposed novel loss functions to handle these problems in the form of min-over-all and min-with-flow losses, both based on the per-pixel minimum reprojection error of Monodepth2 and extended to stereo images and optical flow. With extensive pretraining and novel losses, our model outperformed existing unsupervised approaches in quantitative depth estimation and the ability to distinguish small objects against a background, as evaluated by KITTI 2015.Hyukdoo ChoiIEEEarticleMonocular depth estimationdepth predictionconvolutional neural networksself-supervised learningunsupervised learningElectrical engineering. Electronics. Nuclear engineeringTK1-9971ENIEEE Access, Vol 9, Pp 157236-157246 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	Monocular depth estimation depth prediction convolutional neural networks self-supervised learning unsupervised learning Electrical engineering. Electronics. Nuclear engineering TK1-9971
spellingShingle	Monocular depth estimation depth prediction convolutional neural networks self-supervised learning unsupervised learning Electrical engineering. Electronics. Nuclear engineering TK1-9971 Hyukdoo Choi Self-Supervised Monocular Depth Estimation With Extensive Pretraining
description	Although depth estimation is a key technology for three-dimensional sensing applications involving motion, active sensors such as LiDAR and depth cameras tend to be expensive and bulky. Here, we explore the potential of monocular depth estimation (MDE) based on a self-supervised approach. MDE is a promising technology, but supervised learning suffers from a need for accurate ground-truth depth data. Recent studies have enabled self-supervised training on an MDE model with only monocular image sequences and image-reconstruction errors. We pretrained networks using multiple datasets, including monocular and stereo image sequences. The main challenges posed by the self-supervised MDE model were occlusions and dynamic objects. We proposed novel loss functions to handle these problems in the form of min-over-all and min-with-flow losses, both based on the per-pixel minimum reprojection error of Monodepth2 and extended to stereo images and optical flow. With extensive pretraining and novel losses, our model outperformed existing unsupervised approaches in quantitative depth estimation and the ability to distinguish small objects against a background, as evaluated by KITTI 2015.
format	article
author	Hyukdoo Choi
author_facet	Hyukdoo Choi
author_sort	Hyukdoo Choi
title	Self-Supervised Monocular Depth Estimation With Extensive Pretraining
title_short	Self-Supervised Monocular Depth Estimation With Extensive Pretraining
title_full	Self-Supervised Monocular Depth Estimation With Extensive Pretraining
title_fullStr	Self-Supervised Monocular Depth Estimation With Extensive Pretraining
title_full_unstemmed	Self-Supervised Monocular Depth Estimation With Extensive Pretraining
title_sort	self-supervised monocular depth estimation with extensive pretraining
publisher	IEEE
publishDate	2021
url	https://doaj.org/article/05fc891f131343b9b113a4ba39a130ca
work_keys_str_mv	AT hyukdoochoi selfsupervisedmonoculardepthestimationwithextensivepretraining
_version_	1718403963890434048

Self-Supervised Monocular Depth Estimation With Extensive Pretraining

Ejemplares similares