Research map / Detection and segmentation / CNN features
FreeAnchor: Learning to Match Anchors for Visual Object Detection
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 |
|---|---|---|---|
| Replace RetinaNet's hand-crafted anchor assignment with learning-to-match | RetinaNet, ResNet-50 · COCO val | AP: 35.7 -> 38.7 | Helped |
| Replace RetinaNet's hand-crafted anchor assignment with learning-to-match | RetinaNet, ResNet-101 · COCO val | AP: 37.8 -> 40.9 | Helped |
| Use learning-to-match anchors and detection customized likelihood | RetinaNet, ResNet-50 · COCO val | NR: 81.8 -> 83.8 | Helped |
| Extend the scale range from [640, 800] to [480, 960] | FreeAnchor with ResNeXt-101 · COCO | AP: 44.9 -> 46.0 | Helped |
| Add multi-scale testing over scales {480, 640, 800, 960, 1120, 1280} | FreeAnchor* with ResNeXt-101 · COCO | AP: 46.0 -> 47.3 | Helped |
| Set anchor bag size to 50 rather than the other evaluated sizes (ablation) | FreeAnchor · COCO val | AP | Helped |
| Set background IoU threshold to 0.6 rather than the other evaluated thresholds (ablation) | FreeAnchor · COCO val | AP | Helped |
| Set focal loss parameters to alpha = 0.5 and gamma = 2.0 (ablation) | FreeAnchor · COCO val | AP | Helped |
| Set loss regularization factor beta to 0.75 (ablation) | FreeAnchor · COCO val | AP | Helped |
Closest papers by meaning
- MetaAnchor: Learning to Detect Objects with Customized Anchors (NeurIPS 2018 · similarity 0.90)
- Multiple Anchor Learning for Visual Object Detection (CVPR 2020 · similarity 0.89)
- Bridging the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection (CVPR 2020 · similarity 0.89)
- Order Matters: Competition-Guided Query Ordering for RNN-Based Object Detection (NeurIPS 2026 · similarity 0.84)
- DetKDS: Knowledge Distillation Search for Object Detectors (ICML 2024 · similarity 0.84)
- Improving Single Domain-Generalized Object Detection: A Focus on Diversification and Alignment (CVPR 2024 · 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.