Research map / Detection and segmentation / CNN features
Mask Scoring R-CNN
Zhaojin Huang, Lichao Huang, Yongchao Gong, Chang Huang, Xinggang Wang
Check 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 a MaskIoU head to calibrate mask confidence | Mask R-CNN with ResNet-18/50/101 FPN · COCO 2017 val | AP_m: 27.7 -> 29.3; 34.5 -> 36.0; 36.6 -> 38.2 | Improved |
| Add a MaskIoU head to calibrate mask confidence | Mask R-CNN with ResNet-18/50/101 FPN · COCO 2017 val | AP_m@0.75: 29.0 -> 31.3; 36.7 -> 38.8; 39.0 -> 41.5 | Improved |
| Add a MaskIoU head | Mask R-CNN with ResNet-101 · COCO 2017 val | AP_m: 33.9 -> 35.0; 36.6 -> 38.2; 37.7 -> 39.1 | Improved |
| Add a MaskIoU head and calibrate mask score | Mask R-CNN · COCO 2017 test-dev | AP: 34.3 -> 35.4; 37.0 -> 38.3; 38.4 -> 39.6 | Improved |
| Use target mask concatenated with RoI features rather than target-mask multiplication (ablation) | MaskIoU head · COCO 2017 val | AP: 29.1 -> 29.3 | Improved |
| Use target mask concatenated with RoI features rather than all class masks concatenated with RoI (ablation) | MaskIoU head · COCO 2017 val | AP: 29.1 -> 29.3 | Improved |
| Use target mask with standard RoI features rather than high-resolution RoI features (ablation) | MaskIoU head · COCO 2017 val | AP: 29.1 -> 29.3 | Improved |
| Regress MaskIoU for the target category only instead of all categories, including absent ones (ablation) | MaskIoU head · COCO 2017 val | AP: 24.5 -> 29.3 | Improved |
| Regress MaskIoU for the target category only instead of all positive categories (ablation) | MaskIoU head · COCO 2017 val | AP: 28.2 -> 29.3 | Improved |
| Train on all examples rather than filtering to MaskIoU above 0.3 (ablation) | MaskIoU head · COCO 2017 val | AP: 29.2 -> 29.3 | Improved |
| Train on all examples rather than filtering to MaskIoU above 0.5 (ablation) | MaskIoU head · COCO 2017 val | AP: 29.0 -> 29.3 | Improved |
| Train on all examples rather than filtering to MaskIoU above 0.7 (ablation) | MaskIoU head · COCO 2017 val | AP: 28.8 -> 29.3 | Improved |
Closest papers by meaning
- Mask R-CNN (ICCV 2017 · similarity 0.91)
- Mask Encoding for Single Shot Instance Segmentation (CVPR 2020 · similarity 0.89)
- The Surprising Impact of Mask-Head Architecture on Novel Class Segmentation (ICCV 2021 · similarity 0.88)
- PointINS: Point-Based Instance Segmentation (TPAMI 2021 · similarity 0.87)
- BlendMask: Top-Down Meets Bottom-Up for Instance Segmentation (CVPR 2020 · similarity 0.87)
- MaskDiff: Modeling Mask Distribution with Diffusion Probabilistic Model for Few-Shot Instance Segmentation (AAAI 2024 · 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.
