Research map / Detection and segmentation / CNN features
Projecting Points to Axes: Oriented Object Detection via Point-Axis Representation
Zeyang Zhao, Qilong Xue, Yifan Bai, Yuhang He, Xing Wei, Yihong Gong
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| add point queries with vanilla self-attention to baseline decoder (ablation) | Oriented DETR · DOTA val (R-50) | mAP50 (%): 72.80 -> 70.98 | Hurt |
| use group self-attention within each instance's point queries (ablation) | Oriented DETR · DOTA val (R-50) | mAP50 (%): 70.98 -> 74.21 | Improved |
| decouple deformable cross-attention for center and boundary points (ablation) | Oriented DETR · DOTA val (R-50) | mAP50 (%): 74.21 -> 75.35 | Improved |
| full points detection decoder vs baseline (ablation) | Oriented DETR · DOTA val (R-50) | mAP75: 45.25 -> 50.14 | Improved |
| add top-2 or top-3 max-projection constraints (ablation) | max-projection loss · DOTA val (R-50) | mAP50 (%): 75.35 -> 74.77 (top-2), 73.20 (top-3) | Hurt |
| add penalty for points outside ground-truth box (ablation) | max-projection loss · DOTA val (R-50) | mAP50 (%): 75.35 -> 75.20 | No clear effect |
| increase number of points K from 5 to 13 (ablation) | Oriented DETR · DOTA val (R-50) | mAP50 (%): 74.86 -> 75.35 | Improved |
| increase number of points K beyond 16 to 19 (ablation) | Oriented DETR · DOTA val (R-50) | mAP50 (%): 75.53 -> 75.10 | No clear effect |
| Oriented DETR (point-axis) vs Oriented RepPoints with same backbone | Oriented RepPoints · DOTA (single-scale) | mAP50 (%): 77.6 -> 79.8 | Improved |
| Oriented DETR vs best prior on HRSC2016 VOC07 metric | RTMDet · HRSC2016 | mAP (VOC07) (%): 90.60 -> 90.56 | No clear effect |
| Oriented DETR extended to general detection vs DAB-Deformable-DETR | DAB-Deformable-DETR · COCO val | AP: 49.7 -> 50.5 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ReDet — ReDet: A Rotation-Equivariant Detector for Aerial Object Detection (CVPR 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PointOBB: Learning Oriented Object Detection via Single Point Supervision (CVPR 2024 · similarity 0.89)
- FRED: Towards a Full Rotation-Equivariance in Aerial Image Object Detection (AAAI 2024 · similarity 0.88)
- PointOBB-v2: Towards Simpler, Faster, and Stronger Single Point Supervised Oriented Object Detection (ICLR 2025 · similarity 0.88)
- ABBSPO: Adaptive Bounding Box Scaling and Symmetric Prior based Orientation Prediction for Detecting Aerial Image Objects (CVPR 2025 · similarity 0.87)
- Point2RBox: Combine Knowledge from Synthetic Visual Patterns for End-to-end Oriented Object Detection with Single Point Supervision (CVPR 2024 · similarity 0.86)
- Rethinking Boundary Discontinuity Problem for Oriented Object Detection (CVPR 2024 · similarity 0.86)
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.
