Research map / Detection and segmentation / CNN features
Learn to Pay Attention
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Add two-level parameterised attention with concatenated descriptors to VGG | VGG · CIFAR-10 | Top-1 error: 7.77 (0.08) -> 5.23 (0.04) | Helped |
| Add three-level parameterised attention with concatenated descriptors to VGG | VGG · CIFAR-100 | Top-1 error: 30.62 (0.16) -> 22.97 (0.04) | Helped |
| Add two-level parameterised attention with concatenated descriptors to VGG | VGG · CUB-200-2011 | Top-1 error: 34.64 (0.26) -> 26.80 (0.16) | Helped |
| Add three-level parameterised attention with concatenated descriptors to VGG | VGG · SVHN | Top-1 error: 4.27 (0.04) -> 3.52 (0.04) | Helped |
| Add attention to RN-164 | RN-164 · CIFAR-10 | Top-1 error: 0.9% | Hurt |
| Add attention to RN-164 | RN-164 · CIFAR-100 | Top-1 error: 1.5% | Hurt |
| Add attention at two levels to RN-164 | RN-164 · Caltech-101, Caltech-256, Event-8, Action-40, Scene-67 | Top-1 accuracy | Helped |
| Use three attention levels instead of two in VGG (ablation) | VGG-att · CIFAR-100 | Top-1 error: 24.00 (0.06) -> 22.97 (0.04) | Helped |
| Use dot-product compatibility instead of parameterised compatibility (ablation) | VGG-att · CIFAR-10 | Top-1 error: 5.67 (0.04) -> 6.14 (0.06) | Hurt |
| Use dot-product compatibility instead of parameterised compatibility (ablation) | VGG-att2-concat · CIFAR-100 | Top-1 error: 23.19 (0.04) -> 23.91 (0.11) | Hurt |
| Use independent classifiers rather than concatenating descriptors (ablation) | VGG-att2-dp · CIFAR-100 | Top-1 error: 23.91 (0.11) -> 23.24 (0.07) | Helped |
| Use independent classifiers rather than concatenating descriptors (ablation) | VGG-att2-pc · CIFAR-100 | Top-1 error: 23.19 (0.04) -> 24.00 (0.06) | Hurt |
Closest papers by meaning
- SimAM: A Simple, Parameter-Free Attention Module for Convolutional Neural Networks (ICML 2021 · similarity 0.90)
- Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer (ICLR 2017 · similarity 0.90)
- Residual Attention Network for Image Classification (CVPR 2017 · similarity 0.90)
- Explicitly Modeled Attention Maps for Image Classification (AAAI 2021 · similarity 0.89)
- One Explanation is Not Enough: Structured Attention Graphs for Image Classification (NeurIPS 2021 · similarity 0.88)
- Context-aware Attentional Pooling (CAP) for Fine-grained Visual Classification (AAAI 2021 · similarity 0.88)
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.