Research map / Detection and segmentation / CNN features
SharpContour: A Contour-Based Boundary Refinement Approach for Efficient and Accurate Instance Segmentation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 SharpContour contour refinement | Mask R-CNN* · COCO val2017 | AP: 35.2 -> 37.5 (+2.3) | Helped |
| Add SharpContour contour refinement | DANCE · COCO val2017 | Boundary AP: 20.2 -> 23.9(+3.7) | Helped |
| Add SharpContour contour refinement | CondInst · COCO val2017 | AP: 35.7 -> 37.9 (+2.2) | Helped |
| Add SharpContour to Mask R-CNN | Mask R-CNN · Cityscapes validation | AP: 33.8 -> 37.7(+3.9) | Helped |
| Add SharpContour refinement | RefineMask output masks · COCO val2017 | Boundary AP: 24.7 -> 25.8(+1.1) | Helped |
| Jointly train the segmentation model with SharpContour rather than train SharpContour alone (ablation) | Mask R-CNN + SharpContour · COCO val2017 | AP: 36.2 -> 37.5 | Helped |
| Train Mask R-CNN jointly with SharpContour (ablation) | Mask R-CNN · COCO val2017 | AP: 35.2 -> 35.5 | Helped |
| Use adaptive step size based on IPC uncertainty (ablation) | SharpContour with Mask R-CNN · COCO val2017 | AP: 37.0(-0.5) -> 37.5 | Helped |
| Use 0.0015 rather than the selected deformation ratio 0.003 (ablation) | SharpContour with Mask R-CNN · COCO val2017 | AP: 37.5 -> 37.1 | Hurt |
| Use 0.006 rather than the selected deformation ratio 0.003 (ablation) | SharpContour with Mask R-CNN · COCO val2017 | AP: 37.5 -> 36.6 | Hurt |
| Increase the number of evolution iterations (ablation) | SharpContour with DANCE, Mask R-CNN, or CondInst · COCO val2017 | AP: DANCE: 35.2 -> 36.2; Mask R-CNN: 36.6 -> 37.7; CondInst: 36.9 -> 38.0 | Helped |
| Use regression-based offsets instead of discrete contour movement (ablation) | SharpContour with Mask R-CNN · COCO val2017 | AP: 36.1(+0.9) / 36.2(+1.0) vs 37.5(+2.3) | Hurt |
Closest papers by meaning
- Look Closer To Segment Better: Boundary Patch Refinement for Instance Segmentation (CVPR 2021 · similarity 0.89)
- E2EC: An End-to-End Contour-Based Method for High-Quality High-Speed Instance Segmentation (CVPR 2022 · similarity 0.89)
- PolarNeXt: Rethink Instance Segmentation with Polar Representation (CVPR 2025 · similarity 0.86)
- EFDTR: Learnable Elliptical Fourier Descriptor Transformer for Instance Segmentation (ICML 2025 · similarity 0.85)
- P2Seg: Pointly-supervised Segmentation via Mutual Distillation (ICLR 2024 · similarity 0.85)
- Self-Distilled Depth Refinement with Noisy Poisson Fusion (NeurIPS 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.