Research Article
Triplet Attention Enhanced DeepLab V3+ for Semantic Segmentation: Improving Feature Extraction and Fine-Grained Understanding
@INPROCEEDINGS{10.4108/eai.21-11-2024.2354635, author={Zhuoran Li and Xun Shu and Yancong Deng}, title={Triplet Attention Enhanced DeepLab V3+ for Semantic Segmentation: Improving Feature Extraction and Fine-Grained Understanding}, proceedings={Proceedings of the 2nd International Conference on Machine Learning and Automation, CONF-MLA 2024, November 21, 2024, Adana, Turkey}, publisher={EAI}, proceedings_a={CONF-MLA}, year={2025}, month={3}, keywords={semantic segmentation triplet attention mechanism deeplabv3+ improved feature extraction}, doi={10.4108/eai.21-11-2024.2354635} }- Zhuoran Li
Xun Shu
Yancong Deng
Year: 2025
Triplet Attention Enhanced DeepLab V3+ for Semantic Segmentation: Improving Feature Extraction and Fine-Grained Understanding
CONF-MLA
EAI
DOI: 10.4108/eai.21-11-2024.2354635
Abstract
Semantic segmentation is a critical task in computer vision that requires assigning class labels to individual pixels for a deeper understanding of visual scenes. This paper explores methods to enhance the DeepLab V3+ model by integrating attention mechanisms, specifically Triplet Attention and CBAM, into the backbone, ASPP module, and decoder. We conducted extensive experiments using the Cityscapes dataset to assess the impact of these attention mechanisms. Our results demonstrate that Triplet Attention, particularly when applied to the backbone, significantly improves segmentation performance with minimal computational overhead, outperforming CBAM in most configurations. This study highlights the effectiveness of attention mechanisms in improving the precision and semantic understanding of segmentation models. Future work will explore additional attention mechanisms and expand their application to broader vision-related tasks.
