Research map / Detection and segmentation / Weakly supervised detection and segmentation
OBBSeg: Irregular Lesion Segmentation under Oriented Bounding Box Annotations
Jun Wei, Xinchang Liu, Yu Liu, Chuhua Yang, Shuhui Wang, Hui Huang
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 Mask-to-OBB loss | WeakPolyp · polyp avg (FSPD) | Dice (%): 76.7 -> 79.0 | Improved |
| add M2O loss to OBB-supervised SAM2 base (ablation) | SAM2 base · 5 polyp datasets | Dice (%): 77.2 -> 80.5 | Improved |
| add DBFE without prompts (ablation) | SAM2 base+M2O · 5 polyp datasets | Dice (%): 80.5 -> 83.3 | Improved |
| add PAFE without prompts (ablation) | SAM2 base+M2O · 5 polyp datasets | Dice (%): 80.5 -> 81.3 | Improved |
| add DBFE and PAFE together (ablation) | SAM2 base+M2O · 5 polyp datasets | Dice (%): 80.5 -> 83.6 | Improved |
| add DBFE with prompt supervision (Circle) (ablation) | SAM2 base+M2O+PS · 5 polyp datasets | Dice (%): 82.5 -> 91.8 | Improved |
| add PAFE on top of DBFE (Circle prompt) (ablation) | SAM2 base+M2O+PS+DBFE · 5 polyp datasets | Dice (%): 91.8 -> 94.0 | Improved |
| add PAFE on top of DBFE (Point prompt) (ablation) | SAM2 base+M2O+PS+DBFE · 5 polyp datasets | Dice (%): 81.7 -> 88.5 | Improved |
| use OBB instead of axis-aligned box under matched annotation budget (+Box prompt) | OBBSeg · 5 polyp datasets | Dice (%): 84.7% -> 93.5% | Improved |
| use OBB instead of axis-aligned box under matched budget (+Circle prompt) | OBBSeg · 5 polyp datasets | Dice (%): 93.2% -> 93.7% | No clear effect |
| use OBB supervision rather than box for boundary accuracy (Box prompt) | OBBSeg · Kvasir | HD95: 22.5 -> 6.8 | Improved |
| add M2O to WeakPolyp for elongated lesions | WeakPolyp · FSPD elongated | Dice (%): 65.9 -> 71.2 | Improved |
Closest papers by meaning
- Rethinking Box Supervision: Bias-Free Weakly Supervised Medical Segmentation (CVPR 2026 · similarity 0.87)
- AD-GBC: Anisotropic Granular-Ball Skip-Connection Refiner for UNet-Based Medical Image Segmentation (CVPR 2026 · similarity 0.85)
- EMCAD: Efficient Multi-scale Convolutional Attention Decoding for Medical Image Segmentation (CVPR 2024 · similarity 0.84)
- Why Invariance is Not Enough for Biomedical Domain Generalization and How to Fix It (NeurIPS 2026 · similarity 0.84)
- Interactive Medical Image Segmentation: A Benchmark Dataset and Baseline (CVPR 2025 · similarity 0.84)
- Theoretically Achieving Continuous Representation of Oriented Bounding Boxes (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.
