PointMTL: Multi-Transform Learning for Effective 3D Point Cloud Representations

Effectively learning and extracting the feature representations of 3D point clouds is an important yet challenging task. Most of existing works achieve reasonable performance in 3D vision tasks by modeling the relationships among points appropriately. However, the feature representations are only le...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Yifan Jian, Yuwei Yang, Zhi Chen, Xianguo Qing, Yang Zhao, Liang He, Xuekun Chen, Wei Luo
Formato:	article
Lenguaje:	EN
Publicado:	IEEE 2021
Materias:	3D point clouds feature representations multi-transform learning semantic segmentation Electrical engineering. Electronics. Nuclear engineering TK1-9971
Acceso en línea:	https://doaj.org/article/9e3f67a428074777ae56849b7057d65f
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:9e3f67a428074777ae56849b7057d65f
record_format	dspace
spelling	oai:doaj.org-article:9e3f67a428074777ae56849b7057d65f2021-11-19T00:06:53ZPointMTL: Multi-Transform Learning for Effective 3D Point Cloud Representations2169-353610.1109/ACCESS.2021.3094624https://doaj.org/article/9e3f67a428074777ae56849b7057d65f2021-01-01T00:00:00Zhttps://ieeexplore.ieee.org/document/9474491/https://doaj.org/toc/2169-3536Effectively learning and extracting the feature representations of 3D point clouds is an important yet challenging task. Most of existing works achieve reasonable performance in 3D vision tasks by modeling the relationships among points appropriately. However, the feature representations are only learned with a specific transform through these methods, which are easy to overlap and thus limit the representation ability of the model. To address these issues, we propose a novel Multi-Transform Learning framework for point clouds (PointMTL), which can extract diverse features from multiple mapping transform to obtain richer representations. Specifically, we build a module named Multi-Transform Encoder (MTE), which encodes and aggregates local features from multiple non-linear transforms. To further explore global context representations, a module named Global Spatial Fusion (GSF) is proposed to capture global information and selectively fuse with local representations. Moreover, to guarantee the richness and diversity of learned representations, we further propose a Spatial Independence Criterion (SIC) strategy to enlarge the differences between the transforms and reduce information redundancies. In contrast to previous works, our approach fully exploits representations from multiple transforms, thus having strong expressiveness and good robustness for point clouds related tasks. The experiments on three typical tasks (i.e., semantic segmentation on S3DIS and ScanNet, part segmentation on ShapeNet and shape classification on ModelNet40) demonstrates the effectiveness of our method.Yifan JianYuwei YangZhi ChenXianguo QingYang ZhaoLiang HeXuekun ChenWei LuoIEEEarticle3D point cloudsfeature representationsmulti-transform learningsemantic segmentationElectrical engineering. Electronics. Nuclear engineeringTK1-9971ENIEEE Access, Vol 9, Pp 126241-126255 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	3D point clouds feature representations multi-transform learning semantic segmentation Electrical engineering. Electronics. Nuclear engineering TK1-9971
spellingShingle	3D point clouds feature representations multi-transform learning semantic segmentation Electrical engineering. Electronics. Nuclear engineering TK1-9971 Yifan Jian Yuwei Yang Zhi Chen Xianguo Qing Yang Zhao Liang He Xuekun Chen Wei Luo PointMTL: Multi-Transform Learning for Effective 3D Point Cloud Representations
description	Effectively learning and extracting the feature representations of 3D point clouds is an important yet challenging task. Most of existing works achieve reasonable performance in 3D vision tasks by modeling the relationships among points appropriately. However, the feature representations are only learned with a specific transform through these methods, which are easy to overlap and thus limit the representation ability of the model. To address these issues, we propose a novel Multi-Transform Learning framework for point clouds (PointMTL), which can extract diverse features from multiple mapping transform to obtain richer representations. Specifically, we build a module named Multi-Transform Encoder (MTE), which encodes and aggregates local features from multiple non-linear transforms. To further explore global context representations, a module named Global Spatial Fusion (GSF) is proposed to capture global information and selectively fuse with local representations. Moreover, to guarantee the richness and diversity of learned representations, we further propose a Spatial Independence Criterion (SIC) strategy to enlarge the differences between the transforms and reduce information redundancies. In contrast to previous works, our approach fully exploits representations from multiple transforms, thus having strong expressiveness and good robustness for point clouds related tasks. The experiments on three typical tasks (i.e., semantic segmentation on S3DIS and ScanNet, part segmentation on ShapeNet and shape classification on ModelNet40) demonstrates the effectiveness of our method.
format	article
author	Yifan Jian Yuwei Yang Zhi Chen Xianguo Qing Yang Zhao Liang He Xuekun Chen Wei Luo
author_facet	Yifan Jian Yuwei Yang Zhi Chen Xianguo Qing Yang Zhao Liang He Xuekun Chen Wei Luo
author_sort	Yifan Jian
title	PointMTL: Multi-Transform Learning for Effective 3D Point Cloud Representations
title_short	PointMTL: Multi-Transform Learning for Effective 3D Point Cloud Representations
title_full	PointMTL: Multi-Transform Learning for Effective 3D Point Cloud Representations
title_fullStr	PointMTL: Multi-Transform Learning for Effective 3D Point Cloud Representations
title_full_unstemmed	PointMTL: Multi-Transform Learning for Effective 3D Point Cloud Representations
title_sort	pointmtl: multi-transform learning for effective 3d point cloud representations
publisher	IEEE
publishDate	2021
url	https://doaj.org/article/9e3f67a428074777ae56849b7057d65f
work_keys_str_mv	AT yifanjian pointmtlmultitransformlearningforeffective3dpointcloudrepresentations AT yuweiyang pointmtlmultitransformlearningforeffective3dpointcloudrepresentations AT zhichen pointmtlmultitransformlearningforeffective3dpointcloudrepresentations AT xianguoqing pointmtlmultitransformlearningforeffective3dpointcloudrepresentations AT yangzhao pointmtlmultitransformlearningforeffective3dpointcloudrepresentations AT lianghe pointmtlmultitransformlearningforeffective3dpointcloudrepresentations AT xuekunchen pointmtlmultitransformlearningforeffective3dpointcloudrepresentations AT weiluo pointmtlmultitransformlearningforeffective3dpointcloudrepresentations
_version_	1718420626197184512

PointMTL: Multi-Transform Learning for Effective 3D Point Cloud Representations

Ejemplares similares