Research map / Architectures and efficiency / CNN design
BA-Net: Bridge Attention for Deep Convolutional Neural Networks
Yue Zhao, Junzhou Chen, Zirui Zhang, Ronghui Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| bridge conv1 and conv2 outputs of the current block into the attention layer (ablation) | SE ResNet-50 · ImageNet-1K | top-1 accuracy (%): 78.14 -> 78.85 | Improved |
| bridge the previous block's attention weights instead of the current block's convs (ablation) | BA ResNet-50 · ImageNet-1K | top-1 accuracy (%): 78.85 -> 78.41 | Hurt |
| bridge previous block conv3 instead of current block convs (ablation) | BA ResNet-50 · ImageNet-1K | top-1 accuracy (%): 78.85 -> 78.49 | Hurt |
| bridge previous block end output instead of current block convs (ablation) | BA ResNet-50 · ImageNet-1K | top-1 accuracy (%): 78.85 -> 78.54 | Hurt |
| bridge only current conv1 rather than conv1 and conv2 (ablation) | BA ResNet-50 · ImageNet-1K | top-1 accuracy (%): 78.85 -> 78.78 | Hurt |
| bridge only current conv2 rather than conv1 and conv2 (ablation) | BA ResNet-50 · ImageNet-1K | top-1 accuracy (%): 78.85 -> 78.77 | Hurt |
| add bridge attention (ablation) | ResNeXt-50 · ImageNet-1K | top-1 accuracy (%): 78.77 -> 79.58 | Improved |
| add bridge attention (ablation) | EfficientNet-b0 · ImageNet-1K | top-1 accuracy (%): 70.11 -> 71.70 | Improved |
| add bridge attention (ablation) | MobileNetv3-small · ImageNet-1K | top-1 accuracy (%): 65.87 -> 65.85 | No clear effect |
| add bridge attention (ablation) | MobileNetv3-large · ImageNet-1K | top-1 accuracy (%): 73.29 -> 73.50 | Improved |
| replace SENet attention with BA-Net (ablation) | ResNet-101 · ImageNet-1K | top-1 accuracy (%): 79.41 -> 80.03 | Improved |
| replace SENet attention with BA-Net (ablation) | ResNet-50 Faster R-CNN · COCO2017 | box mAP: 37.7 -> 39.5 | Improved |
Closest papers by meaning
- A^2-Nets: Double Attention Networks (NeurIPS 2018 · similarity 0.86)
- Attention Branch Network: Learning of Attention Mechanism for Visual Explanation (CVPR 2019 · similarity 0.86)
- Weight Excitation: Built-in Attention Mechanisms in Convolutional Neural Networks (ECCV 2020 · similarity 0.86)
- SimAM: A Simple, Parameter-Free Attention Module for Convolutional Neural Networks (ICML 2021 · similarity 0.85)
- AttentionRNN: A Structured Spatial Attention Mechanism (ICCV 2019 · similarity 0.85)
- PartialNet: Compute Less, Perform Better (AAAI 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
