Research map / Detection and segmentation / CNN features
Boundary-preserving Mask R-CNN
Tianheng Cheng, Xinggang Wang, Lichao Huang, Wenyu Liu
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 boundary supervision (BCE) to Mask R-CNN (ablation) | Mask R-CNN ResNet-50-FPN · COCO val2017 (600px) | mask AP: 33.2 -> 33.9 | Improved |
| add M2B and B2M fusion blocks (ablation) | BMask R-CNN · COCO val2017 (600px) | mask AP: 33.9 -> 34.2 | Improved |
| add BCE-Dice loss (ablation) | BMask R-CNN · COCO val2017 (600px) | mask AP: 34.2 -> 34.4 | Improved |
| add P2 28x28 RoI strategy for boundary features (ablation) | BMask R-CNN · COCO val2017 (600px) | mask AP: 34.4 -> 34.7 | Improved |
| supervise boundary head with boundary instead of mask targets (ablation) | BMask R-CNN · COCO val2017 (600px) | mask AP: 33.9 -> 34.7 | Improved |
| supervise boundary head with boundary instead of no supervision (ablation) | BMask R-CNN · COCO val2017 (600px) | mask AP: 34.0 -> 34.7 | Improved |
| add both M2B and B2M fusion vs none (ablation) | BMask R-CNN · COCO val2017 (600px) | mask AP: 33.7 -> 34.7 | Improved |
| add M2B fusion only (ablation) | BMask R-CNN · COCO val2017 (600px) | mask AP: 33.7 -> 34.2 | Improved |
| add B2M fusion only (ablation) | BMask R-CNN · COCO val2017 (600px) | mask AP: 33.7 -> 33.9 | Improved |
| use Dice-BCE instead of BCE loss (ablation) | BMask R-CNN · COCO val2017 (600px) | mask AP: 34.3 -> 34.7 | Improved |
| use Dice-BCE instead of weighted BCE (ablation) | BMask R-CNN · COCO val2017 (600px) | mask AP: 34.1 -> 34.7 | Improved |
| use Dice-BCE instead of Dice loss (ablation) | BMask R-CNN · COCO val2017 (600px) | mask AP: 34.3 -> 34.7 | Improved |
Closest papers by meaning
- Instance Segmentation With Mask-Supervised Polygonal Boundary Transformers (CVPR 2022 · similarity 0.89)
- RefineMask: Towards High-Quality Instance Segmentation With Fine-Grained Features (CVPR 2021 · similarity 0.88)
- Look Closer To Segment Better: Boundary Patch Refinement for Instance Segmentation (CVPR 2021 · similarity 0.88)
- Mask R-CNN (ICCV 2017 · similarity 0.88)
- BlendMask: Top-Down Meets Bottom-Up for Instance Segmentation (CVPR 2020 · similarity 0.87)
- SipMask: Spatial Information Preservation for Fast Image and Video Instance Segmentation (ECCV 2020 · 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.
