Joint Soft–Hard Attention for Self-Supervised Monocular Depth Estimation

In recent years, self-supervised monocular depth estimation has gained popularity among researchers because it uses only a single camera at a much lower cost than the direct use of laser sensors to acquire depth. Although monocular self-supervised methods can obtain dense depths, the estimation accu...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Chao Fan, Zhenyu Yin, Fulong Xu, Anying Chai, Feiqing Zhang
Formato:	article
Lenguaje:	EN
Publicado:	MDPI AG 2021
Materias:	monocular depth estimation self-supervised learning attention mechanism Chemical technology TP1-1185
Acceso en línea:	https://doaj.org/article/004f2d06aaea4edcb6ef81127fdda6f3
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:004f2d06aaea4edcb6ef81127fdda6f3
record_format	dspace
spelling	oai:doaj.org-article:004f2d06aaea4edcb6ef81127fdda6f32021-11-11T19:00:15ZJoint Soft–Hard Attention for Self-Supervised Monocular Depth Estimation10.3390/s212169561424-8220https://doaj.org/article/004f2d06aaea4edcb6ef81127fdda6f32021-10-01T00:00:00Zhttps://www.mdpi.com/1424-8220/21/21/6956https://doaj.org/toc/1424-8220In recent years, self-supervised monocular depth estimation has gained popularity among researchers because it uses only a single camera at a much lower cost than the direct use of laser sensors to acquire depth. Although monocular self-supervised methods can obtain dense depths, the estimation accuracy needs to be further improved for better applications in scenarios such as autonomous driving and robot perception. In this paper, we innovatively combine soft attention and hard attention with two new ideas to improve self-supervised monocular depth estimation: (1) a soft attention module and (2) a hard attention strategy. We integrate the soft attention module in the model architecture to enhance feature extraction in both spatial and channel dimensions, adding only a small number of parameters. Unlike traditional fusion approaches, we use the hard attention strategy to enhance the fusion of generated multi-scale depth predictions. Further experiments demonstrate that our method can achieve the best self-supervised performance both on the standard KITTI benchmark and the Make3D dataset.Chao FanZhenyu YinFulong XuAnying ChaiFeiqing ZhangMDPI AGarticlemonocular depth estimationself-supervised learningattention mechanismChemical technologyTP1-1185ENSensors, Vol 21, Iss 6956, p 6956 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	monocular depth estimation self-supervised learning attention mechanism Chemical technology TP1-1185
spellingShingle	monocular depth estimation self-supervised learning attention mechanism Chemical technology TP1-1185 Chao Fan Zhenyu Yin Fulong Xu Anying Chai Feiqing Zhang Joint Soft–Hard Attention for Self-Supervised Monocular Depth Estimation
description	In recent years, self-supervised monocular depth estimation has gained popularity among researchers because it uses only a single camera at a much lower cost than the direct use of laser sensors to acquire depth. Although monocular self-supervised methods can obtain dense depths, the estimation accuracy needs to be further improved for better applications in scenarios such as autonomous driving and robot perception. In this paper, we innovatively combine soft attention and hard attention with two new ideas to improve self-supervised monocular depth estimation: (1) a soft attention module and (2) a hard attention strategy. We integrate the soft attention module in the model architecture to enhance feature extraction in both spatial and channel dimensions, adding only a small number of parameters. Unlike traditional fusion approaches, we use the hard attention strategy to enhance the fusion of generated multi-scale depth predictions. Further experiments demonstrate that our method can achieve the best self-supervised performance both on the standard KITTI benchmark and the Make3D dataset.
format	article
author	Chao Fan Zhenyu Yin Fulong Xu Anying Chai Feiqing Zhang
author_facet	Chao Fan Zhenyu Yin Fulong Xu Anying Chai Feiqing Zhang
author_sort	Chao Fan
title	Joint Soft–Hard Attention for Self-Supervised Monocular Depth Estimation
title_short	Joint Soft–Hard Attention for Self-Supervised Monocular Depth Estimation
title_full	Joint Soft–Hard Attention for Self-Supervised Monocular Depth Estimation
title_fullStr	Joint Soft–Hard Attention for Self-Supervised Monocular Depth Estimation
title_full_unstemmed	Joint Soft–Hard Attention for Self-Supervised Monocular Depth Estimation
title_sort	joint soft–hard attention for self-supervised monocular depth estimation
publisher	MDPI AG
publishDate	2021
url	https://doaj.org/article/004f2d06aaea4edcb6ef81127fdda6f3
work_keys_str_mv	AT chaofan jointsofthardattentionforselfsupervisedmonoculardepthestimation AT zhenyuyin jointsofthardattentionforselfsupervisedmonoculardepthestimation AT fulongxu jointsofthardattentionforselfsupervisedmonoculardepthestimation AT anyingchai jointsofthardattentionforselfsupervisedmonoculardepthestimation AT feiqingzhang jointsofthardattentionforselfsupervisedmonoculardepthestimation
_version_	1718431624921612288

Joint Soft–Hard Attention for Self-Supervised Monocular Depth Estimation

Ejemplares similares