TCANet: three-stream coordinate attention network for RGB-D indoor semantic segmentation
Abstract Semantic segmentation plays a vital role in indoor scene analysis. Currently, its accuracy is still limited due to the complex conditions of various indoor scenes. In addition, it is difficult to complete this task solely relying on RGB images. Since depth images can provide additional 3D g...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Springer
2023-08-01
|
Series: | Complex & Intelligent Systems |
Subjects: | |
Online Access: | https://doi.org/10.1007/s40747-023-01210-4 |
_version_ | 1797272237815889920 |
---|---|
author | Weikuan Jia Xingchao Yan Qiaolian Liu Ting Zhang Xishang Dong |
author_facet | Weikuan Jia Xingchao Yan Qiaolian Liu Ting Zhang Xishang Dong |
author_sort | Weikuan Jia |
collection | DOAJ |
description | Abstract Semantic segmentation plays a vital role in indoor scene analysis. Currently, its accuracy is still limited due to the complex conditions of various indoor scenes. In addition, it is difficult to complete this task solely relying on RGB images. Since depth images can provide additional 3D geometric information to RGB images, researchers chose to incorporate depth images for improving the accuracy of indoor semantic segmentation. However, it is still a challenge to effectively fuse the depth information with the RGB images. To address this issue, a three-stream coordinate attention network is proposed. The presented network reconstructs a multi-modal feature fusion module for RGB-D features, which can realize the aggregation of two modal information along the spatial and channel dimensions. Meanwhile, three convolutional neural network branches are used to construct a parallel three-stream structure, which can, respectively, process the RGB features, depth features and combined features. On one hand, the proposed network can preserve the original RGB and depth feature streams, simultaneously. On the other hand, it can also contribute to utilize and propagate the fusion feature flow better. The embedded ASPP module is used to optimize the semantic information in the proposed network, so as to aggregate the feature information of different scales and obtain more accurate features. Experimental results show that the proposed model can reach a state-of-the-art mIoU accuracy of 50.2% on the NYUDv2 dataset and on the more complex SUN-RGBD dataset. |
first_indexed | 2024-03-07T14:25:27Z |
format | Article |
id | doaj.art-ac6003e4396e4fb28e64b07f93093fc7 |
institution | Directory Open Access Journal |
issn | 2199-4536 2198-6053 |
language | English |
last_indexed | 2024-03-07T14:25:27Z |
publishDate | 2023-08-01 |
publisher | Springer |
record_format | Article |
series | Complex & Intelligent Systems |
spelling | doaj.art-ac6003e4396e4fb28e64b07f93093fc72024-03-06T08:07:01ZengSpringerComplex & Intelligent Systems2199-45362198-60532023-08-011011219123010.1007/s40747-023-01210-4TCANet: three-stream coordinate attention network for RGB-D indoor semantic segmentationWeikuan Jia0Xingchao Yan1Qiaolian Liu2Ting Zhang3Xishang Dong4School of Information Science and Engineering, Zaozhuang UniversitySchool of Information Science and Engineering, Shandong Normal UniversitySchool of Information Science and Engineering, Zaozhuang UniversitySchool of Information Science and Engineering, Zaozhuang UniversitySchool of Information Science and Engineering, Zaozhuang UniversityAbstract Semantic segmentation plays a vital role in indoor scene analysis. Currently, its accuracy is still limited due to the complex conditions of various indoor scenes. In addition, it is difficult to complete this task solely relying on RGB images. Since depth images can provide additional 3D geometric information to RGB images, researchers chose to incorporate depth images for improving the accuracy of indoor semantic segmentation. However, it is still a challenge to effectively fuse the depth information with the RGB images. To address this issue, a three-stream coordinate attention network is proposed. The presented network reconstructs a multi-modal feature fusion module for RGB-D features, which can realize the aggregation of two modal information along the spatial and channel dimensions. Meanwhile, three convolutional neural network branches are used to construct a parallel three-stream structure, which can, respectively, process the RGB features, depth features and combined features. On one hand, the proposed network can preserve the original RGB and depth feature streams, simultaneously. On the other hand, it can also contribute to utilize and propagate the fusion feature flow better. The embedded ASPP module is used to optimize the semantic information in the proposed network, so as to aggregate the feature information of different scales and obtain more accurate features. Experimental results show that the proposed model can reach a state-of-the-art mIoU accuracy of 50.2% on the NYUDv2 dataset and on the more complex SUN-RGBD dataset.https://doi.org/10.1007/s40747-023-01210-4Depth imagesIndoor semantic segmentationThree-streamCoordinate attention |
spellingShingle | Weikuan Jia Xingchao Yan Qiaolian Liu Ting Zhang Xishang Dong TCANet: three-stream coordinate attention network for RGB-D indoor semantic segmentation Complex & Intelligent Systems Depth images Indoor semantic segmentation Three-stream Coordinate attention |
title | TCANet: three-stream coordinate attention network for RGB-D indoor semantic segmentation |
title_full | TCANet: three-stream coordinate attention network for RGB-D indoor semantic segmentation |
title_fullStr | TCANet: three-stream coordinate attention network for RGB-D indoor semantic segmentation |
title_full_unstemmed | TCANet: three-stream coordinate attention network for RGB-D indoor semantic segmentation |
title_short | TCANet: three-stream coordinate attention network for RGB-D indoor semantic segmentation |
title_sort | tcanet three stream coordinate attention network for rgb d indoor semantic segmentation |
topic | Depth images Indoor semantic segmentation Three-stream Coordinate attention |
url | https://doi.org/10.1007/s40747-023-01210-4 |
work_keys_str_mv | AT weikuanjia tcanetthreestreamcoordinateattentionnetworkforrgbdindoorsemanticsegmentation AT xingchaoyan tcanetthreestreamcoordinateattentionnetworkforrgbdindoorsemanticsegmentation AT qiaolianliu tcanetthreestreamcoordinateattentionnetworkforrgbdindoorsemanticsegmentation AT tingzhang tcanetthreestreamcoordinateattentionnetworkforrgbdindoorsemanticsegmentation AT xishangdong tcanetthreestreamcoordinateattentionnetworkforrgbdindoorsemanticsegmentation |