Viewpoint invariant semantic object and scene categorization with RGB-D sensors

Understanding the semantics of objects and scenes using multi-modal RGB-D sensors serves many robotics applications. Key challenges for accurate RGB-D image recognition are the scarcity of training data, variations due to viewpoint changes and the heterogeneous nature of the data. We address these p...

Full description

Bibliographic Details
Main Authors: Mohd Zaki, Hasan Firdaus, Shafait, Faisal, Mian, Ajmal
Format: Article
Language:English
English
English
Published: Springer New York LLC 2019
Subjects:
Online Access:http://irep.iium.edu.my/64696/
http://irep.iium.edu.my/64696/20/64696_Viewpoint%20invariant%20semantic%20object%20and%20scene_complete.pdf
http://irep.iium.edu.my/64696/19/64696_Viewpoint%20invariant%20semantic%20object%20and%20scene_scopus.pdf
http://irep.iium.edu.my/64696/31/64696_Viewpoint%20invariant%20semantic%20object%20and%20scene%20categorization%20with%20RGB-D%20sensors_WOS.pdf
_version_ 1848786261098102784
author Mohd Zaki, Hasan Firdaus
Shafait, Faisal
Mian, Ajmal
author_facet Mohd Zaki, Hasan Firdaus
Shafait, Faisal
Mian, Ajmal
author_sort Mohd Zaki, Hasan Firdaus
building IIUM Repository
collection Online Access
description Understanding the semantics of objects and scenes using multi-modal RGB-D sensors serves many robotics applications. Key challenges for accurate RGB-D image recognition are the scarcity of training data, variations due to viewpoint changes and the heterogeneous nature of the data. We address these problems and propose a generic deep learning framework based on a pre-trained convolutional neural network, as a feature extractor for both the colour and depth channels. We propose a rich multi-scale feature representation, referred to as convolutional hypercube pyramid (HP-CNN), that is able to encode discriminative information from the convolutional tensors at different levels of detail. We also present a technique to fuse the proposed HP-CNN with the activations of fully connected neurons based on an extreme learning machine classifier in a late fusion scheme which leads to a highly discriminative and compact representation. To further improve performance, we devise HP-CNN-T which is a view-invariant descriptor extracted from a multi-view 3D object pose (M3DOP) model. M3DOP is learned from over 140,000 RGB-D images that are synthetically generated by rendering CAD models from different viewpoints. Extensive evaluations on four RGB-D object and scene recognition datasets demonstrate that our HP-CNN and HP-CNN-T consistently outperforms state-of-the-art methods for several recognition tasks by a significant margin.
first_indexed 2025-11-14T17:06:13Z
format Article
id iium-64696
institution International Islamic University Malaysia
institution_category Local University
language English
English
English
last_indexed 2025-11-14T17:06:13Z
publishDate 2019
publisher Springer New York LLC
recordtype eprints
repository_type Digital Repository
spelling iium-646962019-08-01T02:25:54Z http://irep.iium.edu.my/64696/ Viewpoint invariant semantic object and scene categorization with RGB-D sensors Mohd Zaki, Hasan Firdaus Shafait, Faisal Mian, Ajmal Q300 Cybernetics QA75 Electronic computers. Computer science Understanding the semantics of objects and scenes using multi-modal RGB-D sensors serves many robotics applications. Key challenges for accurate RGB-D image recognition are the scarcity of training data, variations due to viewpoint changes and the heterogeneous nature of the data. We address these problems and propose a generic deep learning framework based on a pre-trained convolutional neural network, as a feature extractor for both the colour and depth channels. We propose a rich multi-scale feature representation, referred to as convolutional hypercube pyramid (HP-CNN), that is able to encode discriminative information from the convolutional tensors at different levels of detail. We also present a technique to fuse the proposed HP-CNN with the activations of fully connected neurons based on an extreme learning machine classifier in a late fusion scheme which leads to a highly discriminative and compact representation. To further improve performance, we devise HP-CNN-T which is a view-invariant descriptor extracted from a multi-view 3D object pose (M3DOP) model. M3DOP is learned from over 140,000 RGB-D images that are synthetically generated by rendering CAD models from different viewpoints. Extensive evaluations on four RGB-D object and scene recognition datasets demonstrate that our HP-CNN and HP-CNN-T consistently outperforms state-of-the-art methods for several recognition tasks by a significant margin. Springer New York LLC 2019-04-01 Article PeerReviewed application/pdf en http://irep.iium.edu.my/64696/20/64696_Viewpoint%20invariant%20semantic%20object%20and%20scene_complete.pdf application/pdf en http://irep.iium.edu.my/64696/19/64696_Viewpoint%20invariant%20semantic%20object%20and%20scene_scopus.pdf application/pdf en http://irep.iium.edu.my/64696/31/64696_Viewpoint%20invariant%20semantic%20object%20and%20scene%20categorization%20with%20RGB-D%20sensors_WOS.pdf Mohd Zaki, Hasan Firdaus and Shafait, Faisal and Mian, Ajmal (2019) Viewpoint invariant semantic object and scene categorization with RGB-D sensors. Autonomous Robots, 43 (4). pp. 1005-1022. ISSN 0929-5593 E-ISSN 1573-7527 (In Press) https://link.springer.com/journal/10514 10.1007/s10514-018-9776-8
spellingShingle Q300 Cybernetics
QA75 Electronic computers. Computer science
Mohd Zaki, Hasan Firdaus
Shafait, Faisal
Mian, Ajmal
Viewpoint invariant semantic object and scene categorization with RGB-D sensors
title Viewpoint invariant semantic object and scene categorization with RGB-D sensors
title_full Viewpoint invariant semantic object and scene categorization with RGB-D sensors
title_fullStr Viewpoint invariant semantic object and scene categorization with RGB-D sensors
title_full_unstemmed Viewpoint invariant semantic object and scene categorization with RGB-D sensors
title_short Viewpoint invariant semantic object and scene categorization with RGB-D sensors
title_sort viewpoint invariant semantic object and scene categorization with rgb-d sensors
topic Q300 Cybernetics
QA75 Electronic computers. Computer science
url http://irep.iium.edu.my/64696/
http://irep.iium.edu.my/64696/
http://irep.iium.edu.my/64696/
http://irep.iium.edu.my/64696/20/64696_Viewpoint%20invariant%20semantic%20object%20and%20scene_complete.pdf
http://irep.iium.edu.my/64696/19/64696_Viewpoint%20invariant%20semantic%20object%20and%20scene_scopus.pdf
http://irep.iium.edu.my/64696/31/64696_Viewpoint%20invariant%20semantic%20object%20and%20scene%20categorization%20with%20RGB-D%20sensors_WOS.pdf