Research map / Detection and segmentation / CNN features
PolyTransform: Deep Polygon Transformer 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 |
|---|---|---|---|
| Refine the UPSNet instance masks into deformable polygons with PolyTransform | UPSNet · Cityscapes val | AP: +3.0 | Helped |
| Refine the UPSNet instance masks into deformable polygons with PolyTransform | UPSNet · Cityscapes val | AF: +10.3 | Helped |
| Refine the UPSNet instance masks into deformable polygons with PolyTransform | UPSNet · New self-driving dataset validation | AP: +2.2 | Helped |
| Refine the UPSNet instance masks into deformable polygons with PolyTransform | UPSNet · New self-driving dataset validation | AF: +5.6 | Helped |
| Refine the Cityscapes Stretch segmentation initialization into polygons | FCN and DeepLabV3 initializations · Cityscapes Stretch | mean IoU: +0.08 to +0.17 | Helped |
| Refine the Cityscapes Stretch segmentation initialization into polygons | FCN and DeepLabV3 initializations · Cityscapes Stretch | F1 at 1px: +1.14 to +1.53 | Helped |
| Refine the Cityscapes Hard segmentation output into polygons | interactive annotation methods · Cityscapes Hard | mean IoU: +4.9% | Helped |
| Use PolyTransform output as the starting annotation for labelers | manual annotation · New self-driving dataset | annotation time: 60.3 minutes -> 39.4 minutes; 35% | Helped |
| Replace polygon deformation with a convolutional mask-refinement network | Cityscapes val refinement of UPSNet · Cityscapes val | AP: 43.8 -> 44.6 | Hurt |
| Replace polygon deformation with an extra UPSNet mask-refinement head | Cityscapes val refinement of UPSNet · Cityscapes val | AP: 43.5 -> 44.6 | Hurt |
Closest papers by meaning
- Instance Segmentation With Mask-Supervised Polygonal Boundary Transformers (CVPR 2022 · similarity 0.91)
- Instance Segmentation by Jointly Optimizing Spatial Embeddings and Clustering Bandwidth (CVPR 2019 · similarity 0.87)
- Training-Free Dataset Pruning for Instance Segmentation (ICLR 2025 · similarity 0.85)
- Any3DIS: Class-Agnostic 3D Instance Segmentation by 2D Mask Tracking (CVPR 2025 · similarity 0.85)
- Insightful Instance Features for 3D Instance Segmentation (CVPR 2025 · similarity 0.84)
- Open3DIS: Open-Vocabulary 3D Instance Segmentation with 2D Mask Guidance (CVPR 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.