Research map / Detection and segmentation / CNN features
AABO: Adaptive Anchor Box Optimization for Object Detection via Bayesian Sub-sampling
Wenshuo Ma, Tingzhong Tian, Hang Xu, Yimin Huang, Zhenguo Li
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| optimize anchor settings (scales, ratios) (ablation) | Faster-RCNN w FPN · COCO | mAP: 36.4 -> 37.1 | Improved |
| optimize RPN head architecture (ablation) | Faster-RCNN w FPN · COCO | mAP: 36.4 -> 36.3 | No clear effect |
| optimize anchors and architecture jointly instead of anchors only (ablation) | Faster-RCNN w FPN · COCO | mAP: 37.1 -> 37.2 | No clear effect |
| replace pre-defined anchors with AABO searched anchors | Faster-RCNN w FPN · COCO | mAP: 36.4 -> 38.8 | Improved |
| replace pre-defined anchors with AABO searched anchors | Faster-RCNN w FPN · VG1000 | mAP: 6.5 -> 8.0 | Improved |
| replace pre-defined anchors with AABO searched anchors | Faster-RCNN w FPN · ADE | mAP: 10.3 -> 11.9 | Improved |
| replace default anchors with searched anchors | DCNv2 w x101 · COCO (SOTA detectors) | mAP: 43.4 -> 45.8 | Improved |
| add Bayesian optimization with Hyperband to random search (ablation) | Anchor search (single space) · COCO | mAP of optimal anchor: 37.2 -> 37.8 | Improved |
| use sub-sampling instead of Hyperband in AABO (ablation) | AABO (single space) · COCO | mAP of optimal anchor: 37.8 -> 38.3 | Improved |
| use feature-map-wise search space instead of single space (ablation) | AABO w SS · COCO | mAP of optimal anchor: 38.3 -> 38.8 | Improved |
| use random search in the feature-map-wise space instead of AABO (ablation) | Random Search · COCO | mAP: 11.5 vs 38.8 | Hurt |
| use K-Means anchors instead of pre-defined anchors | Faster-RCNN w FPN · COCO | mAP: 36.4 -> 37.0 | Improved |
Closest papers by meaning
- Bridging the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection (CVPR 2020 · similarity 0.90)
- Probabilistic Anchor Assignment with IoU Prediction for Object Detection (ECCV 2020 · similarity 0.89)
- Soft Anchor-Point Object Detection (ECCV 2020 · similarity 0.88)
- Anchor-Intermediate Detector: Decoupling and Coupling Bounding Boxes for Accurate Object Detection (ICCV 2023 · similarity 0.88)
- ObjectBox: From Centers to Boxes for Anchor-Free Object Detection (ECCV 2022 · similarity 0.88)
- ABBSPO: Adaptive Bounding Box Scaling and Symmetric Prior based Orientation Prediction for Detecting Aerial Image Objects (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.
