Research map / Detection and segmentation / Weakly supervised detection and segmentation
PointOBB-v3: Expanding Performance Boundaries of Single Point-Supervised Oriented Object Detection
Peiyuan Zhang, Junwei Luo, Xue Yang, Yi Yu, Qingyun Li, Yue Zhou, Xiaosong Jia, Xudong Lü, Jingdong Chen, Xiang Li, Junchi Yan, Yansheng Li
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 SSC loss (ablation) | PointOBB-v3 two-stage · DIOR-R (Rotated FCOS) | mAP50: 30.16 -> 36.39 | Improved |
| add DS matching (ablation) | PointOBB-v3 two-stage · DIOR-R (Rotated FCOS) | mAP50: 36.39 -> 37.31 | Improved |
| add SSFF module (ablation) | PointOBB-v3 two-stage · DIOR-R (Rotated FCOS) | mAP50: 37.31 -> 40.18 | Improved |
| add SSFF module (ablation) | PointOBB-v3 two-stage · DOTA-v1.0 (Oriented R-CNN) | mAP50: 33.31 -> 50.44 | Improved |
| add resized and rot/flp views to MIL loss (ablation) | PointOBB-v3 two-stage · DIOR-R | mAP50: 34.94 -> 38.08 | Improved |
| use gradual instead of plain burn-in (ablation) | PointOBB-v3 two-stage · DIOR-R | mAP50: 23.71 -> 32.23 | Improved |
| attach angle branch gradients to base network (ablation) | PointOBB-v3 two-stage · DIOR-R | mAP50: 32.23 -> 38.08 | Improved |
| group scores by scale instead of proposal (ablation) | PointOBB-v3 two-stage · DIOR-R | mAP50: 30.42 -> 38.08 | Improved |
| group scores by scale instead of ratio (ablation) | PointOBB-v3 two-stage · DIOR-R | mAP50: 36.71 -> 38.08 | Improved |
| use 10% noisy point labels instead of center points (ablation) | PointOBB-v3 two-stage · DIOR-R | mAP50: 32.11 -> 38.08 | Improved |
| widen point range from 10% to 20% (ablation) | PointOBB-v3 two-stage · DIOR-R | mAP50: 38.08 -> 35.78 | Hurt |
| burn-in steps 6 and 8 versus 0 and 0 (ablation) | PointOBB-v3 two-stage · DIOR-R | mAP50: 38.33 -> 41.82 | Improved |
Closest papers by meaning
- PointOBB-v2: Towards Simpler, Faster, and Stronger Single Point Supervised Oriented Object Detection (ICLR 2025 · similarity 0.95)
- PointOBB: Learning Oriented Object Detection via Single Point Supervision (CVPR 2024 · similarity 0.94)
- Point2RBox: Combine Knowledge from Synthetic Visual Patterns for End-to-end Oriented Object Detection with Single Point Supervision (CVPR 2024 · similarity 0.90)
- Point2RBox-v2: Rethinking Point-supervised Oriented Object Detection with Spatial Layout Among Instances (CVPR 2025 · similarity 0.89)
- PointSR: Self-Regularized Point Supervision for Drone-View Object Detection (CVPR 2025 · similarity 0.87)
- ABBSPO: Adaptive Bounding Box Scaling and Symmetric Prior based Orientation Prediction for Detecting Aerial Image Objects (CVPR 2025 · 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.
