Research map / Detection and segmentation / CNN features
Look Closer To Segment Better: Boundary Patch Refinement for 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 boundary patch refinement | Mask R-CNN* · Cityscapes test | AP: 32.6 -> 36.9 | Helped |
| Add BPR after SegFix | PolyTransform + SegFix · Cityscapes test | AP: 41.2 -> 42.7 | Helped |
| Transfer BPR trained on Mask R-CNN to another segmentation model | PointRend · Cityscapes | AF: 58.0 -> 66.5 | Helped |
| Add BPR refinement | Mask R-CNN ResNeXt-FPN-101 · COCO val2017 | AP: 38.4 -> 39.2 | Helped |
| Add binary mask patches to image patches (ablation) | BPR · Cityscapes val | AP: 20.1 -> 39.8 | Helped |
| Use 64×64 patches without padding instead of larger or differently padded patches (ablation) | BPR · Cityscapes val | AP: 39.4; 39.7; 39.8; 39.7; 39.6 | Mixed |
| Extract boundary patches by dense sampling and NMS rather than predefined grid or whole-instance crops (ablation) | BPR · Cityscapes val | AP: 39.8 vs 39.3, 39.1, 38.8, 37.5, 38.7 | Helped |
| Increase refinement-network input size from 64 to 256 (ablation) | BPR · Cityscapes val | AP: 39.1 -> 40.0 | Helped |
| Increase refinement backbone capacity from HRNet-W18s to HRNet-W48 (ablation) | BPR · Cityscapes val | AP: 39.8 -> 40.1 | Helped |
| Increase the inference NMS threshold and retain more overlapping patches (ablation) | BPR · Cityscapes val | AP: 37.7 -> 40.1 | Helped |
| Train BPR on PointRend or SegFix predictions and transfer it to Mask R-CNN | Mask R-CNN · Cityscapes val | AP: 36.4 -> 38.7; 36.4 -> 39.1 | Helped |
Closest papers by meaning
- Instance Segmentation With Mask-Supervised Polygonal Boundary Transformers (CVPR 2022 · similarity 0.91)
- RefineMask: Towards High-Quality Instance Segmentation With Fine-Grained Features (CVPR 2021 · similarity 0.90)
- PatchDCT: Patch Refinement for High Quality Instance Segmentation (ICLR 2023 · similarity 0.90)
- P2Seg: Pointly-supervised Segmentation via Mutual Distillation (ICLR 2024 · similarity 0.85)
- The Devil is in the Object Boundary: Towards Annotation-free Instance Segmentation using Foundation Models (ICLR 2024 · similarity 0.85)
- Training-Free Dataset Pruning for Instance Segmentation (ICLR 2025 · similarity 0.85)
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.