Monocular Human Depth Estimation Via Pose Estimation

We propose a novel monocular depth estimator, which improves the prediction accuracy on human regions by utilizing pose information. The proposed algorithm consists of two networks — PoseNet and DepthNet — to estimate keypoint heatmaps and a depth map, respectively. We incorpor...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Jinyoung Jun, Jae-Han Lee, Chul Lee, Chang-Su Kim
Formato:	article
Lenguaje:	EN
Publicado:	IEEE 2021
Materias:	Monocular depth estimation human pose estimation human depth estimation loss rebalancing strategy Electrical engineering. Electronics. Nuclear engineering TK1-9971
Acceso en línea:	https://doaj.org/article/362374746b434f51bfc40a80c1a6f80d
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:362374746b434f51bfc40a80c1a6f80d
record_format	dspace
spelling	oai:doaj.org-article:362374746b434f51bfc40a80c1a6f80d2021-11-17T00:00:17ZMonocular Human Depth Estimation Via Pose Estimation2169-353610.1109/ACCESS.2021.3126629https://doaj.org/article/362374746b434f51bfc40a80c1a6f80d2021-01-01T00:00:00Zhttps://ieeexplore.ieee.org/document/9606929/https://doaj.org/toc/2169-3536We propose a novel monocular depth estimator, which improves the prediction accuracy on human regions by utilizing pose information. The proposed algorithm consists of two networks — PoseNet and DepthNet — to estimate keypoint heatmaps and a depth map, respectively. We incorporate the pose information from PoseNet to improve the depth estimation performance of DepthNet. Specifically, we develop the feature blending block, which fuses the features from PoseNet and DepthNet and feeds them into the next layer of DepthNet, to make the networks learn to predict the depths of human regions more accurately. Furthermore, we develop a novel joint training scheme using partially labeled datasets, which balances multiple loss functions effectively by adjusting weights. Experimental results demonstrate that the proposed algorithm can improve depth estimation performance significantly, especially around human regions. For example, the proposed algorithm improves the depth estimation performance on the human regions of ResNet-50 by 2.8% and 7.0% in terms of <inline-formula> <tex-math notation="LaTeX">$\delta _{1}$ </tex-math></inline-formula> and RMSE, respectively, on the proposed HD + P dataset.Jinyoung JunJae-Han LeeChul LeeChang-Su KimIEEEarticleMonocular depth estimationhuman pose estimationhuman depth estimationloss rebalancing strategyElectrical engineering. Electronics. Nuclear engineeringTK1-9971ENIEEE Access, Vol 9, Pp 151444-151457 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	Monocular depth estimation human pose estimation human depth estimation loss rebalancing strategy Electrical engineering. Electronics. Nuclear engineering TK1-9971
spellingShingle	Monocular depth estimation human pose estimation human depth estimation loss rebalancing strategy Electrical engineering. Electronics. Nuclear engineering TK1-9971 Jinyoung Jun Jae-Han Lee Chul Lee Chang-Su Kim Monocular Human Depth Estimation Via Pose Estimation
description	We propose a novel monocular depth estimator, which improves the prediction accuracy on human regions by utilizing pose information. The proposed algorithm consists of two networks — PoseNet and DepthNet — to estimate keypoint heatmaps and a depth map, respectively. We incorporate the pose information from PoseNet to improve the depth estimation performance of DepthNet. Specifically, we develop the feature blending block, which fuses the features from PoseNet and DepthNet and feeds them into the next layer of DepthNet, to make the networks learn to predict the depths of human regions more accurately. Furthermore, we develop a novel joint training scheme using partially labeled datasets, which balances multiple loss functions effectively by adjusting weights. Experimental results demonstrate that the proposed algorithm can improve depth estimation performance significantly, especially around human regions. For example, the proposed algorithm improves the depth estimation performance on the human regions of ResNet-50 by 2.8% and 7.0% in terms of <inline-formula> <tex-math notation="LaTeX">$\delta _{1}$ </tex-math></inline-formula> and RMSE, respectively, on the proposed HD + P dataset.
format	article
author	Jinyoung Jun Jae-Han Lee Chul Lee Chang-Su Kim
author_facet	Jinyoung Jun Jae-Han Lee Chul Lee Chang-Su Kim
author_sort	Jinyoung Jun
title	Monocular Human Depth Estimation Via Pose Estimation
title_short	Monocular Human Depth Estimation Via Pose Estimation
title_full	Monocular Human Depth Estimation Via Pose Estimation
title_fullStr	Monocular Human Depth Estimation Via Pose Estimation
title_full_unstemmed	Monocular Human Depth Estimation Via Pose Estimation
title_sort	monocular human depth estimation via pose estimation
publisher	IEEE
publishDate	2021
url	https://doaj.org/article/362374746b434f51bfc40a80c1a6f80d
work_keys_str_mv	AT jinyoungjun monocularhumandepthestimationviaposeestimation AT jaehanlee monocularhumandepthestimationviaposeestimation AT chullee monocularhumandepthestimationviaposeestimation AT changsukim monocularhumandepthestimationviaposeestimation
_version_	1718426040341102592

Monocular Human Depth Estimation Via Pose Estimation

Ejemplares similares