Research map / Detection and segmentation / CNN features
Region Proposal by Guided Anchoring
Check 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 |
|---|---|---|---|
| Replace dense uniform anchors with guided location and shape prediction | RPN · MS COCO 2017 val | AR1000: 59.4 -> 68.5 | Helped |
| Replace sliding-window anchors with guided anchoring | Fast R-CNN · MS COCO 2017 test-dev | AP: 37.1 -> 39.4 | Helped |
| Replace sliding-window anchors with guided anchoring | Faster R-CNN · MS COCO 2017 test-dev | AP: 37.1 -> 39.8 | Helped |
| Replace sliding-window anchors with guided anchoring | RetinaNet · MS COCO 2017 test-dev | AP: 35.9 -> 37.1 | Helped |
| Fine-tune the trained detector using GA-RPN proposals | Faster R-CNN · MS COCO 2017 test-dev | AP: 37.4 -> 39.6 | Helped |
| Add location prediction to the otherwise standard RPN (ablation) | RPN · MS COCO 2017 val | AR1000: 59.4 -> 59.5 | No clear effect |
| Add shape prediction (ablation) | RPN · MS COCO 2017 val | AR1000: 59.4 -> 63.6 | Helped |
| Add feature adaptation conditioned on predicted anchor shapes (ablation) | GA-RPN · MS COCO 2017 val | AR1000: 63.8 -> 68.5 | Helped |
| Condition deformable-convolution offsets on anchor shape predictions (ablation) | GA-RPN · MS COCO 2017 val | AR1000: 66.1 -> 68.5 | Helped |
| Keep anchor centers aligned with feature pixels instead of predicting center offsets (ablation) | GA-RPN · MS COCO 2017 val | AR1000: 63.8 -> 68.5 | Helped |
| Increase location threshold from 0 to 0.01 (ablation) | GA-RPN · MS COCO 2017 val | AR1000: 68.5 -> 68.5 | No clear effect |
| Increase location threshold to 0.05 (ablation) | GA-RPN · MS COCO 2017 val | AR1000: 68.5 -> 68.2 | Hurt |
Closest papers by meaning
- Multiple Anchor Learning for Visual Object Detection (CVPR 2020 · similarity 0.89)
- Dynamic Anchor Learning for Arbitrary-Oriented Object Detection (AAAI 2021 · similarity 0.88)
- MetaAnchor: Learning to Detect Objects with Customized Anchors (NeurIPS 2018 · similarity 0.87)
- On the Use of Anchoring for Training Vision Models (NeurIPS 2024 · similarity 0.86)
- Recognize Any Regions (NeurIPS 2024 · similarity 0.86)
- BOOTPLACE: Bootstrapped Object Placement with Detection Transformers (CVPR 2025 · similarity 0.84)
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.