Optics and Precision Engineering, Volume. 33, Issue 1, 123(2025)
Improved DeepLabv3+ semantic segmentation incorporating attention mechanisms
Figures & Tables(11)
Fig. 6. Visualisation of experiment results of original network and improved Deeplabv3+
Fig. 7. mIoU and mPA of DeepLabv3+ segmentation network before improvement
Table 1. ASPP module improvement comparison experiment results
View tableView in ArticleTable 1. ASPP module improvement comparison experiment results
网络模型 Backbone mIoU mPA mPrecision mRecall Deeplabv3+ MobileNetV2 69.77% 82.15% 83.61% 82.15% DeepLabv3+_HWD MobileNetV2 70.63% 82.39% 84.63% 82.05%
Table 2. Comparative experiment results of different attention mechanisms
View tableView in ArticleTable 2. Comparative experiment results of different attention mechanisms
Backbone Attention mIoU mPA mPrecision mRecall MobileNetV2_HWD ECA 71.31% 82.56% 84.79% 82.17% MobileNetV2_HWD SimAM 71.64% 82.70% 84.14% 82.55% MobileNetV2_HWD EANet 71.87% 82.86% 83.79% 82.52%
Table 3. Experimental accuracy results for different module combinations
View tableView in ArticleTable 3. Experimental accuracy results for different module combinations
MobileNetV2 SimAM ECA EANet HWD mIoU mPA mPrecision mRecall √ 69.77% 82.15% 83.61% 82.15% √ √ √ 71.64% 82.70% 84.14% 82.55% √ √ √ 71.31% 82.56% 84.79% 82.17% √ √ √ 71.87% 82.86% 83.79% 82.52% √ √ √ √ 72.59% 83.01% 83.52% 83.01%
Tools
Get Citation
Copy Citation Text
He YAN, Qiuxia LEI, Xu WANG. Improved DeepLabv3+ semantic segmentation incorporating attention mechanisms[J]. Optics and Precision Engineering, 2025, 33(1): 123
Paper Information
Category:
Received: Jun. 28, 2024
Accepted: --
Published Online: Apr. 1, 2025
The Author Email: He YAN (yanhe@ cqut.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20