Research map / Detection and segmentation / Open-set and few-shot detection
Learning what and where to attend
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 ClickMe map supervision to GALA-ResNet-50 | GALA-ResNet-50 · ClickMe test split | top-1 error: 53.90 -> 49.29 | Helped |
| Add ClickMe map supervision to GALA-ResNet-50 | GALA-ResNet-50 · ClickMe test split | maps: 64.21$^{**}$ -> 88.56$^{**}$ | Helped |
| Replace SE attention with GALA attention | ResNet-50 · ILSVRC12 validation | top-1 error: 23.26 -> 22.73 | Helped |
| Add ClickMe map supervision on available training images | GALA-ResNet-50 · Full ILSVRC12 training; validation | top-1 error: 27.28 -> 26.17 | Helped |
| Add ClickMe map supervision to GALA-ResNet-50 | GALA-ResNet-50 · COCO 2017 detection validation subset | IoU: 0.03 -> 0.26 | Helped |
| Replace detailed ClickMe maps with bounding boxes (ablation) | GALA-ResNet-50 w/ ClickMe · ClickMe test split | top-1 error: 49.29 -> 58.14 | Hurt |
| Replace detailed ClickMe maps with bounding boxes (ablation) | GALA-ResNet-50 w/ ClickMe · Full ILSVRC12 training; validation | top-1 error: 26.17 -> 27.24 | Hurt |
| Apply ClickMe supervision directly to feature activity without an attention module (ablation) | ResNet-50 · ClickMe test split | top-1 error: 63.68 -> 61.32 | No clear effect |
| Tune the ClickMe-loss weight to λ=6 (ablation) | GALA-ResNet-50 · ClickMe validation | object categorization accuracy | Helped |
Closest papers by meaning
- Harmonizing the object recognition strategies of deep neural networks with humans (NeurIPS 2022 · similarity 0.89)
- Bi-Directional Object-Context Prioritization Learning for Saliency Ranking (CVPR 2022 · similarity 0.88)
- Tell Me Where to Look: Guided Attention Inference Network (CVPR 2018 · similarity 0.88)
- Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer (ICLR 2017 · similarity 0.88)
- OverLoCK: An Overview-first-Look-Closely-next ConvNet with Context-Mixing Dynamic Kernels (CVPR 2025 · similarity 0.86)
- L-WISE: Boosting Human Visual Category Learning Through Model-Based Image Selection and Enhancement (ICLR 2025 · similarity 0.86)
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.