Research map / Detection and segmentation / CNN features
PointRend: Image Segmentation As Rendering
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 |
|---|---|---|---|
| Replace the standard 4× conv mask head with PointRend | Mask R-CNN with ResNet-50-FPN · COCO val2017 | mask AP: 35.2 -> 36.3 | Helped |
| Replace the standard 4× conv mask head with PointRend | Mask R-CNN with ResNet-50-FPN · COCO categories evaluated on LVIS annotations | AP*: 37.6 -> 39.7 | Helped |
| Replace the standard 4× conv mask head with PointRend | Mask R-CNN with ResNet-50-FPN · Cityscapes instance segmentation | mask AP: 33.0 -> 35.8 | Helped |
| Use adaptive subdivision instead of dense 4× conv prediction at the same 224×224 resolution | Mask R-CNN mask head · 224×224 instance masks | FLOPs: 34B -> 0.9B | Helped |
| Increase PointRend output resolution from 28×28 to 224×224 (ablation) | PointRend on Mask R-CNN · COCO val2017 | mask AP: 36.1 -> 36.3 | Helped |
| Increase PointRend output resolution from 28×28 to 224×224 (ablation) | PointRend on Mask R-CNN · COCO categories evaluated on LVIS annotations | AP*: 39.2 -> 39.7 | Helped |
| Increase PointRend output resolution from 28×28 to 224×224 (ablation) | PointRend on Mask R-CNN · Cityscapes instance segmentation | mask AP: 35.4 -> 35.8 | Helped |
| Increase points per subdivision step from 14² to 28² (ablation) | PointRend at 224×224 · COCO val2017 | mask AP: 36.1 -> 36.3 | Helped |
| Increase points per subdivision step from 28² to 56² (ablation) | PointRend at 224×224 · COCO val2017 | mask AP: 36.3 -> 36.3 | No clear effect |
| Bias training samples mildly toward uncertain regions while retaining uniform coverage (ablation) | PointRend · COCO val2017 | mask AP: 35.9 -> 36.3 | Helped |
| Bias training samples heavily toward uncertain regions (ablation) | PointRend · COCO val2017 | mask AP: 35.9 -> 34.4 | Hurt |
| Add PointRend to DeepLabV3-OS-16 | DeepLabV3 semantic segmentation · Cityscapes | mIoU: 77.2 -> 78.4 | Helped |
Closest papers by meaning
- Pointly-Supervised Instance Segmentation (CVPR 2022 · similarity 0.88)
- FreePoint: Unsupervised Point Cloud Instance Segmentation (CVPR 2024 · similarity 0.86)
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · similarity 0.86)
- BFANet: Revisiting 3D Semantic Segmentation with Boundary Feature Analysis (CVPR 2025 · similarity 0.86)
- Point-SAM: Promptable 3D Segmentation Model for Point Clouds (ICLR 2025 · similarity 0.85)
- A Generative Approach at the Instance-Level for Image Segmentation Under Limited Training Data Conditions (Student Abstract) (AAAI 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.