Research map / Detection and segmentation / CNN features
Multiple Anchor Learning for Visual Object Detection
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 fixed anchor assignment with joint anchor selection for classification and localization | RetinaNet with ResNet-50 · MS-COCO test-dev | AP: 35.5 -> 39.2 | Helped |
| Replace fixed anchor assignment with joint anchor selection for classification and localization | RetinaNet with ResNet-101 · MS-COCO test-dev | AP: 39.1 -> 43.6 | Helped |
| Replace fixed anchor assignment with joint anchor selection for classification and localization | RetinaNet with ResNeXt-101 · MS-COCO test-dev | AP: 40.8 -> 45.9 | Helped |
| Use anchor bags instead of scattered anchors (ablation) | RetinaNet, ResNet-50 · COCO-minval | AP: 35.46 -> 38.14 | Helped |
| Gradually reduce selected anchors from all to top-1 (ablation) | MAL+S(all), ResNet-50 · COCO-minval | AP: 38.14 -> 38.39 | Helped |
| Use 40 instead of 50 anchors per bag (ablation) | MAL, ResNet-50 · COCO-minval | AP: 38.39 -> 38.27 | No clear effect |
| Use 60 instead of 50 anchors per bag (ablation) | MAL, ResNet-50 · COCO-minval | AP: 38.39 -> 38.08 | No clear effect |
| Add constant-rate anchor depression (ablation) | RetinaNet, ResNet-50 · COCO-minval | AP: 35.46 -> 35.25 | Hurt |
| Use step-scheduled instead of constant-rate anchor depression (ablation) | RetinaNet, ResNet-50 · COCO-minval | AP: 35.46 -> 35.88 | Helped |
| Use symmetric step-scheduled anchor depression (ablation) | RetinaNet, ResNet-50 · COCO-minval | AP: 35.46 -> 36.18 | Helped |
| Combine anchor selection and depression adversarially (ablation) | RetinaNet, ResNet-50 · COCO-minval | AP: 35.5% vs. 39.2% | Helped |
| Use multi-scale testing | MAL with ResNet-101 · MS-COCO test-dev | AP: 43.6 -> 45.0 | Helped |
Closest papers by meaning
- Region Proposal by Guided Anchoring (CVPR 2019 · similarity 0.89)
- FreeAnchor: Learning to Match Anchors for Visual Object Detection (NeurIPS 2019 · similarity 0.89)
- Bridging the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection (CVPR 2020 · similarity 0.89)
- MetaAnchor: Learning to Detect Objects with Customized Anchors (NeurIPS 2018 · similarity 0.89)
- Feature Selective Anchor-Free Module for Single-Shot Object Detection (CVPR 2019 · similarity 0.88)
- Adaptive Object Detection Using Adjacency and Zoom Prediction (CVPR 2016 · similarity 0.87)
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.