A channel attention feature recalibration method based on low-rank space interaction
By adopting a channel attention feature recalibration method based on low-rank spatial interaction, the shortcomings of channel attention in terms of model complexity, accuracy and real-time performance are addressed, achieving a better balance and making it applicable to mainstream vision models.
Patent Information
- Application Number
- CN202610348582.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-20
- Publication Date
- 2026-07-17
AI Technical Summary
Existing channel attention methods fall short of ideal performance in terms of model complexity, accuracy, and real-time performance, and suffer from problems such as information loss and parameter redundancy.
We adopt a channel attention feature recalibration method based on low-rank spatial interaction. We retain global statistical information through compound pooling operation, use one-dimensional convolutional layers to learn the interaction relationship between channels, and only take the first 3 to 9 weights for forward propagation during the inference stage to reduce model complexity.
It improves the model's feature representation ability, reduces the number of model parameters, and enhances the balance between real-time performance and accuracy. It is applicable to mainstream vision models such as ImageNet, YOLO, and Mask R-CNN.
Smart Images
Figure FT_1 
Figure FT_2 
Figure FT_3