Coordinate Attention for Efficient Mobile Network Design
By Qibin Hou, Daquan Zhou, Jiashi Feng. 2021
5,022 citations recorded by the source corpus.
Recent studies on mobile network design have demonstrated the remarkable effectiveness of channel attention (e.g., the Squeeze-and-Excitation attention) for lifting model performance, but they generally neglect the positional information, which is important for generating spatially selective attention maps. In this paper, we propose a novel attention mechanism for mobile networks by embedding positional information into channel attention, which we call "coordinate attention". Unlike channel attention that transforms a feature tensor to a single feature vector via 2D global pooling, the coordinate attention factorizes channel attention into two 1D feature encoding processes that aggregate features along the two spatial directions, respectively. In this way, long-range dependencies can be captured along one spatial direction and meanwhile precise positional information can be preserved along the other spatial direction. The resulting feature maps are then encoded separately into a pair of direction-aware and position-sensitive attention maps that can be complementarily applied to the input feature map to augment the representations of the objects of interest. Our coordinate attention is simple and can be flexibly plugged into classic mobile networks, such as MobileNetV2, MobileNeXt, and EfficientNet with nearly no computational overhead. Extensive experiments demonstrate that our coordinate attention is not only beneficial to ImageNet classification but more interestingly, behaves better in down-stream tasks, such as object detection and semantic segmentation. Code is available at https://github.com/Andrew-Qibin/CoordAttention.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Electrical and Electronic Engineering, Model Architectures, Efficiency & Compression
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
In Computer Vision and Pattern Recognition
- Jiashi Feng
- Shuicheng Yan
- Dacheng Tao
- Ming-Ming Cheng
- Luc Van Gool
- Qibin Hou
- Ming-Hsuan Yang
- Yi Yang
In Electrical and Electronic Engineering
In Model Architectures
In Efficiency & Compression
Citations in the corpus
- DeepViT: Towards Deeper Vision Transformer (2021, background): “The recent progress of CNNs largely benefits from novel architecture design strategies that make training deep CNNs more effective [9, 35, 13, 12, 36, 52].”
- Review of Image Classification Algorithms Based on Convolutional Neural Networks (2021, methodology): “[150] proposed the Coordinate Attention (CA) module for efficient mobile networks, which uses two 1D GAP to capture feature codes in two spatial directions (positional information), CA block as shown in Figure 29.”
- Slim-neck by GSConv: a lightweight-design for real-time detector architectures (2022, background): “Moreover, the attention mechanisms, such as SPP [26], SE [27], CBAM [28] and CA [29], could improve the efficiency and performance, especially for the lightweight detectors [10-15].”
- FcaNet: Frequency Channel Attention Networks (2020, methodology): “Inspired by these works, a series of works like BAM [28], DAN [9], CBAM [39], scSE [32], and CoordAtten-tion [19] are proposed to fuse spatial attention [45] and channel attention.”
- Constructing Stronger and Faster Baselines for Skeleton-Based Action Recognition (2021, background): “attention [25], we propose a novel attention module, named Spatial Temporal Joint Attention (ST-JointAtt), to jointly distinguish the most informative joints in certain frames from the whole skeleton sequence.”
- DS-TransUNet: Dual Swin Transformer U-Net for Medical Image Segmentation (2021, background): “There exists some works trying to model long-range dependencies for convolution such as attention mechanism [9] [10] [11].”
- GhostNetV2: Enhance Cheap Operation with Long-Range Attention (2022, methodology): “SE [14] and CBAM [38] are two widely-used attention modules, and CA [11] is a SOTA method presented recently.”
- SCSA: Exploring the Synergistic Effects Between Spatial and Channel Attention (2024, neutral): “Existing plug-and-play attention methods can be primarily categorized into three types: channel attention [23, 27, 62, 55, 31, 41], spatial attention [22, 56, 42, 9], and hybrid channel-spatial attention [57, 40, 38, 14, 20, 65, 39, 61].”
- Vision Permutator: A Permutable MLP-Like Architecture for Visual Recognition (2021, methodology): “Different from concurrent popular MLP-like models listed in Table 2, we separately encoding the token representations along the height and width dimensions, generating position-sensitive outputs that are crucial for locating and identifying objects of interest (Hou et al., 2021; Wang et al., 2020).”
- Distract Your Attention: Multi-Head Cross Attention Network for Facial Expression Recognition (2021, background): “[23] propose a coordinate attention that embeds positional information into channel attention to generate spatially selective attention maps.”
- ASF-YOLO: A Novel YOLO Model with Attentional Scale Sequence Fusion for Cell Instance Segmentation (2023, methodology): “Compared with the channel attention SENet, the channel and spatial attention CBAM, and the spatial attention Coordinate Attention (CA) [53], the proposed CPAM attention mechanism provides better performance despite with slight increase in computation amount and parameters, as shown in Table 4.”
- Slim-neck by GSConv: A better design paradigm of detector architectures for autonomous vehicles (2022, background): “More, the attention mechanisms, such as SPP [25], SE [26], CBAM [27] and CA [28], can improve the efficiency and performance of detectors, especially for the lightweight detectors.”
- Attention mechanisms in computer vision: A survey (2021, background): “[130, 101, 123, 86, 96, 17, 84], [108, 40, 61, 166, 148, 152, 135] , [122, 14, 19, 62, 78]”
- Efficient Multi-Scale Attention Module with Cross-Spatial Learning (2023, methodology): “Furthermore, the MobileNetv2 network with the CBAM module significantly improves the Top-1 validation accuracy performance over the baseline model by 1.3%.”