Efficient Attention Calibration Network for Real-Time Semantic Segmentation

Hengfeng Zha, Rui Liu, Dongsheng Zhou, Xin Yang, Qiang Zhang, Xiaopeng Wei
Proceedings of The 12th Asian Conference on Machine Learning, PMLR 129:337-352, 2020.

Abstract

In recent years, the attention mechanism has been widely used in computer vision. Semantic segmentation, as one of the fundamental tasks of computer vision, has been subject to tremendous development as a result. But because of its huge computing overhead, attention-based approaches are difficult to use for real-time applications such as self-driving. In this paper, we propose a self-calibration method baesd on self-attentiion that successfully applies the attention mechanism to real-time semantic segmentation. Specifically, a spatial attention module to adjust the edges of the coarse segmentation results which gained from the real-time semantic segmentation backbone network, and obtain more granular segmentation results. We refer to this method as the Efficient Attentional Calibration Network (EACNet). Experiments on the Cityscapes dataset validate the efficiency and performance of the method. With the high-resolution input and without any post-processing, EACNet achieved 72.4% mIoU of accuracy while running at 116.9 FPS. Compared to other state-of-the-art methods for real-time semantic segmentation, our network gained a better balance between performance and speed.

Cite this Paper

BibTeX
@InProceedings{pmlr-v129-zha20a, title = {Efficient Attention Calibration Network for Real-Time Semantic Segmentation}, author = {Zha, Hengfeng and Liu, Rui and Zhou, Dongsheng and Yang, Xin and Zhang, Qiang and Wei, Xiaopeng}, booktitle = {Proceedings of The 12th Asian Conference on Machine Learning}, pages = {337--352}, year = {2020}, editor = {Pan, Sinno Jialin and Sugiyama, Masashi}, volume = {129}, series = {Proceedings of Machine Learning Research}, month = {18--20 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v129/zha20a/zha20a.pdf}, url = {https://proceedings.mlr.press/v129/zha20a.html}, abstract = {In recent years, the attention mechanism has been widely used in computer vision. Semantic segmentation, as one of the fundamental tasks of computer vision, has been subject to tremendous development as a result. But because of its huge computing overhead, attention-based approaches are difficult to use for real-time applications such as self-driving. In this paper, we propose a self-calibration method baesd on self-attentiion that successfully applies the attention mechanism to real-time semantic segmentation. Specifically, a spatial attention module to adjust the edges of the coarse segmentation results which gained from the real-time semantic segmentation backbone network, and obtain more granular segmentation results. We refer to this method as the Efficient Attentional Calibration Network (EACNet). Experiments on the Cityscapes dataset validate the efficiency and performance of the method. With the high-resolution input and without any post-processing, EACNet achieved 72.4% mIoU of accuracy while running at 116.9 FPS. Compared to other state-of-the-art methods for real-time semantic segmentation, our network gained a better balance between performance and speed.} }
Endnote
%0 Conference Paper %T Efficient Attention Calibration Network for Real-Time Semantic Segmentation %A Hengfeng Zha %A Rui Liu %A Dongsheng Zhou %A Xin Yang %A Qiang Zhang %A Xiaopeng Wei %B Proceedings of The 12th Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Sinno Jialin Pan %E Masashi Sugiyama %F pmlr-v129-zha20a %I PMLR %P 337--352 %U https://proceedings.mlr.press/v129/zha20a.html %V 129 %X In recent years, the attention mechanism has been widely used in computer vision. Semantic segmentation, as one of the fundamental tasks of computer vision, has been subject to tremendous development as a result. But because of its huge computing overhead, attention-based approaches are difficult to use for real-time applications such as self-driving. In this paper, we propose a self-calibration method baesd on self-attentiion that successfully applies the attention mechanism to real-time semantic segmentation. Specifically, a spatial attention module to adjust the edges of the coarse segmentation results which gained from the real-time semantic segmentation backbone network, and obtain more granular segmentation results. We refer to this method as the Efficient Attentional Calibration Network (EACNet). Experiments on the Cityscapes dataset validate the efficiency and performance of the method. With the high-resolution input and without any post-processing, EACNet achieved 72.4% mIoU of accuracy while running at 116.9 FPS. Compared to other state-of-the-art methods for real-time semantic segmentation, our network gained a better balance between performance and speed.
APA
Zha, H., Liu, R., Zhou, D., Yang, X., Zhang, Q. & Wei, X.. (2020). Efficient Attention Calibration Network for Real-Time Semantic Segmentation. Proceedings of The 12th Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 129:337-352 Available from https://proceedings.mlr.press/v129/zha20a.html.

Related Material