FcaNet: Frequency Channel Attention Networks
Zequn Qin, Pengyi Zhang, Fei Wu, Xi Li
Abstract
Attention mechanism, especially channel attention, has gained great success in the computer vision field. Many works focus on how to design efficient channel attention mechanisms while ignoring a fundamental problem, i.e., channel attention mechanism uses scalar to represent channel, which is difficult due to massive information loss. In this work, we start from a different view and regard the channel representation problem as a compression process using frequency analysis. Based on the frequency analysis, we mathematically prove that the conventional global average pooling is a special case of the feature decomposition in the frequency domain. With the proof, we naturally generalize the compression of the channel attention mechanism in the frequency domain and propose our method with multi-spectral channel attention, termed as FcaNet. FcaNet is simple but effective. We can change a few lines of code in the calculation to implement our method within existing channel attention methods. Moreover, the proposed method achieves state-of-the-art results compared with other channel attention methods on image classification, object detection, and instance segmentation tasks. Our method could consistently outperform the baseline SENet, with the same number of parameters and the same computational cost. Our code and models will are publicly available at https: //github.com/cfzd/FcaNet .
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
54 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
Cited by top-tier papers54
- Detecting Camouflaged Object in Frequency DomainYijie Zhong, Bo Li, Lv Tang, Senyun Kuang et al.CVPR 2022 · 271 citations
- Dynamic High-Pass Filtering and Multi-Spectral Attention for Image Super-ResolutionSalma Abdel Magid, Yulun Zhang, Donglai Wei, Won-Dong Jang et al.ICCV 2021 · 122 citations
- IRNeXt: Rethinking Convolutional Network Design for Image RestorationYuning Cui, Wenqi Ren, Sining Yang, Xiaochun Cao et al.ICML 2023 · 112 citations
- Online Convolutional ReparameterizationMu Hu, Junyi Feng, Jiashen Hua, Baisheng Lai et al.CVPR 2022 · 90 citations
- Modality-Agnostic Domain Generalizable Medical Image Segmentation by Multi-Frequency in Multi-Scale AttentionJu-Hyeon Nam, Nur Suriza Syazwany, Su Jung Kim, Sang-Chul LeeCVPR 2024 · 73 citations
Builds on7
- Attention Augmented Convolutional NetworksIrwan Bello, Barret Zoph, Quoc Le, Ashish Vaswani et al.ICCV 2019 · 1,149 citations
- An Empirical Study of Spatial Attention Mechanisms in Deep NetworksXizhou Zhu, Dazhi Cheng, Zheng Zhang, Stephen Lin et al.ICCV 2019 · 522 citations
- SRM: A Style-Based Recalibration Module for Convolutional Neural NetworksHyunJae Lee, Hyo-Eun Kim, Hyeonseob NamICCV 2019 · 286 citations
- Deep Residual Learning in the JPEG Transform DomainMax Ehrlich, Larry DavisICCV 2019 · 145 citations
- Coordinate Attention for Efficient Mobile Network DesignQibin Hou, Daquan Zhou, Jiashi FengCVPR 2021
Related papers
- ECA-Net: Efficient Channel Attention for Deep Convolutional Neural NetworksQilong Wang, Banggu Wu, Pengfei Zhu, Peihua Li et al.CVPR 2020
- MCA: Moment Channel Attention NetworksYangbo Jiang, Zhiwei Jiang, Le Han, Zenan Huang et al.AAAI 2024 · 15 citations
- Fully Attentional Network for Semantic SegmentationQi Song, Jie Li, Chenghong Li, Hao Guo et al.AAAI 2022 · 63 citations
- Squeeze-and-Attention Networks for Semantic SegmentationZilong Zhong, Zhong Qiu Lin, Rene Bidart, Xiaodan Hu et al.CVPR 2020
- Fast and Accurate Lane Detection via Frequency Domain LearningYulin He, Wei Chen, Zhengfa Liang, Dan Chen et al.ACM MM 2021 · 7 citations
