LASNet: A Light-Weight Asymmetric Spatial Feature Network for Real-Time Semantic Segmentation
In recent years, deep learning models have achieved great success in the field of semantic segmentation, which achieve satisfactory performance by introducing a large number of parameters. However, this achievement usually leads to high computational complexity, which seriously limits the deployment...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-10-01
|
| Series: | Electronics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2079-9292/11/19/3238 |
| _version_ | 1797479718147063808 |
|---|---|
| author | Yu Chen Weida Zhan Yichun Jiang Depeng Zhu Renzhong Guo Xiaoyu Xu |
| author_facet | Yu Chen Weida Zhan Yichun Jiang Depeng Zhu Renzhong Guo Xiaoyu Xu |
| author_sort | Yu Chen |
| collection | DOAJ |
| description | In recent years, deep learning models have achieved great success in the field of semantic segmentation, which achieve satisfactory performance by introducing a large number of parameters. However, this achievement usually leads to high computational complexity, which seriously limits the deployment of semantic segmented applications on mobile devices with limited computing and storage resources. To address this problem, we propose a lightweight asymmetric spatial feature network (LASNet) for real-time semantic segmentation. We consider the network parameters, inference speed, and performance to design the structure of LASNet, which can make the LASNet applied to embedded devices and mobile devices better. In the encoding part of LASNet, we propose the LAS module, which retains and utilize spatial information. This module uses a combination of asymmetric convolution, group convolution, and dual-stream structure to reduce the number of network parameters and maintain strong feature extraction ability. In the decoding part of LASNet, we propose the multivariate concatenate module to reuse the shallow features, which can improve the segmentation accuracy and maintain a high inference speed. Our network attains precise real-time segmentation results in a wide range of experiments. Without additional processing and pre-training, LASNet achieves 70.99% mIoU and 110.93 FPS inference speed in the CityScapes dataset with only 0.8 M model parameters. |
| first_indexed | 2024-03-09T21:49:49Z |
| format | Article |
| id | doaj.art-5db64ea871394806a2fc9ae13e60197a |
| institution | Directory Open Access Journal |
| issn | 2079-9292 |
| language | English |
| last_indexed | 2024-03-09T21:49:49Z |
| publishDate | 2022-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Electronics |
| spelling | doaj.art-5db64ea871394806a2fc9ae13e60197a2023-11-23T20:08:42ZengMDPI AGElectronics2079-92922022-10-011119323810.3390/electronics11193238LASNet: A Light-Weight Asymmetric Spatial Feature Network for Real-Time Semantic SegmentationYu Chen0Weida Zhan1Yichun Jiang2Depeng Zhu3Renzhong Guo4Xiaoyu Xu5National Demonstration Center for Experimental Electrical, School of Electronic and Information Engineering, Changchun University of Science and Technology, Changchun 130022, ChinaNational Demonstration Center for Experimental Electrical, School of Electronic and Information Engineering, Changchun University of Science and Technology, Changchun 130022, ChinaNational Demonstration Center for Experimental Electrical, School of Electronic and Information Engineering, Changchun University of Science and Technology, Changchun 130022, ChinaNational Demonstration Center for Experimental Electrical, School of Electronic and Information Engineering, Changchun University of Science and Technology, Changchun 130022, ChinaNational Demonstration Center for Experimental Electrical, School of Electronic and Information Engineering, Changchun University of Science and Technology, Changchun 130022, ChinaNational Demonstration Center for Experimental Electrical, School of Electronic and Information Engineering, Changchun University of Science and Technology, Changchun 130022, ChinaIn recent years, deep learning models have achieved great success in the field of semantic segmentation, which achieve satisfactory performance by introducing a large number of parameters. However, this achievement usually leads to high computational complexity, which seriously limits the deployment of semantic segmented applications on mobile devices with limited computing and storage resources. To address this problem, we propose a lightweight asymmetric spatial feature network (LASNet) for real-time semantic segmentation. We consider the network parameters, inference speed, and performance to design the structure of LASNet, which can make the LASNet applied to embedded devices and mobile devices better. In the encoding part of LASNet, we propose the LAS module, which retains and utilize spatial information. This module uses a combination of asymmetric convolution, group convolution, and dual-stream structure to reduce the number of network parameters and maintain strong feature extraction ability. In the decoding part of LASNet, we propose the multivariate concatenate module to reuse the shallow features, which can improve the segmentation accuracy and maintain a high inference speed. Our network attains precise real-time segmentation results in a wide range of experiments. Without additional processing and pre-training, LASNet achieves 70.99% mIoU and 110.93 FPS inference speed in the CityScapes dataset with only 0.8 M model parameters.https://www.mdpi.com/2079-9292/11/19/3238asymmetric convolutionreal-time semantic segmentationattention mechanismresidual unit |
| spellingShingle | Yu Chen Weida Zhan Yichun Jiang Depeng Zhu Renzhong Guo Xiaoyu Xu LASNet: A Light-Weight Asymmetric Spatial Feature Network for Real-Time Semantic Segmentation Electronics asymmetric convolution real-time semantic segmentation attention mechanism residual unit |
| title | LASNet: A Light-Weight Asymmetric Spatial Feature Network for Real-Time Semantic Segmentation |
| title_full | LASNet: A Light-Weight Asymmetric Spatial Feature Network for Real-Time Semantic Segmentation |
| title_fullStr | LASNet: A Light-Weight Asymmetric Spatial Feature Network for Real-Time Semantic Segmentation |
| title_full_unstemmed | LASNet: A Light-Weight Asymmetric Spatial Feature Network for Real-Time Semantic Segmentation |
| title_short | LASNet: A Light-Weight Asymmetric Spatial Feature Network for Real-Time Semantic Segmentation |
| title_sort | lasnet a light weight asymmetric spatial feature network for real time semantic segmentation |
| topic | asymmetric convolution real-time semantic segmentation attention mechanism residual unit |
| url | https://www.mdpi.com/2079-9292/11/19/3238 |
| work_keys_str_mv | AT yuchen lasnetalightweightasymmetricspatialfeaturenetworkforrealtimesemanticsegmentation AT weidazhan lasnetalightweightasymmetricspatialfeaturenetworkforrealtimesemanticsegmentation AT yichunjiang lasnetalightweightasymmetricspatialfeaturenetworkforrealtimesemanticsegmentation AT depengzhu lasnetalightweightasymmetricspatialfeaturenetworkforrealtimesemanticsegmentation AT renzhongguo lasnetalightweightasymmetricspatialfeaturenetworkforrealtimesemanticsegmentation AT xiaoyuxu lasnetalightweightasymmetricspatialfeaturenetworkforrealtimesemanticsegmentation |