Research map / Detection and segmentation / CNN features
Beyond Bounding-Box: Convex-Hull Feature Adaptation for Oriented and Densely Packed Object Detection
Zonghao Guo, Chang Liu, Xiaosong Zhang, Jianbin Jiao, Xiangyang Ji, Qixiang Ye
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 |
|---|---|---|---|
| use convex-hull feature adaptation | object detection · DOTA | mAP: 76.67% vs 70.70% | Improved |
| use convex-hull feature adaptation | object detection · SKU110K-R | mAP: 57.0% vs. 55.9% | Improved |
| replace Smoothed-L1 loss with CIoU loss (ablation) | Baseline-R · DOTA validation | mAP: 65.35% vs 66.30% | Improved |
| add convex-hull generation (ablation) | Baseline-R · DOTA validation | mAP: +1.88% | Improved |
| add convex-hull set splitting (ablation) | CGen · DOTA validation | mAP: +1.52% | Improved |
| add an anti-aliasing coefficient (ablation) | CGen with set splitting · DOTA validation | mAP: +0.43% | Improved |
| use CIoU loss instead of Smoothed-L1 loss (ablation) | object localization · DOTA validation | mAP: 66.30% vs 65.35% | Improved |
| set the CIoU assignment threshold to 0.2 (ablation) | CFA convex-hull assignment · DOTA validation | mAP | Improved |
| assign six convex-hulls per feature pyramid level to each set (ablation) | CFA convex-hull set construction · DOTA validation | mAP | Improved |
| use anti-aliasing factor γ=0.75 (ablation) | CFA feature adaptation · DOTA validation | mAP | Improved |
Closest papers by meaning
- Gliding Vertex on the Horizontal Bounding Box for Multi-Oriented Object Detection (TPAMI 2020 · similarity 0.85)
- Robust Domain Adaptive Object Detection With Unified Multi-Granularity Alignment (TPAMI 2024 · similarity 0.85)
- Multi-Granularity Alignment Domain Adaptation for Object Detection (CVPR 2022 · similarity 0.85)
- Adaptive Object Detection Using Adjacency and Zoom Prediction (CVPR 2016 · similarity 0.84)
- Union-over-Intersections: Object Detection beyond Winner-Takes-All (ICLR 2025 · similarity 0.83)
- ABBSPO: Adaptive Bounding Box Scaling and Symmetric Prior based Orientation Prediction for Detecting Aerial Image Objects (CVPR 2025 · similarity 0.83)
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.
