Research map / Detection and segmentation / Open-set and few-shot detection
Annotating Object Instances With a Polygon-RNN
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 |
|---|---|---|---|
| Predict object outlines as sequential polygons rather than pixel masks | Cityscapes instance annotation · Cityscapes validation split, prediction mode | Mean IoU (%): 60.21 -> 61.40 | Helped |
| Use Polygon-RNN rather than SharpMask | Cityscapes instance annotation · Cityscapes car | IoU (%): 65.17 -> 71.17 | Helped |
| Use Polygon-RNN rather than SharpMask | Cityscapes instance annotation · Cityscapes bus and train | IoU (%): Bus: 73.02 -> 69.53; Train: 64.06 -> 53.74 | Hurt |
| Add annotator vertex corrections at threshold T=4 | Polygon-RNN on Cityscapes car instances · Cityscapes car instances | IoU: 73.3 -> 82.2 | Helped |
| Allow more corrections by using threshold T=1 instead of T=4 | Polygon-RNN on Cityscapes car instances · Cityscapes car instances | IoU: 82.2 -> 87.7 | Helped |
| Raise correction threshold from 1 to 4, reducing required corrections | Polygon-RNN on Cityscapes all classes · Cityscapes all classes | Mean IOU: 84.74 -> 75.79 | Hurt |
| Train on Cityscapes and apply the model to an unseen dataset | Polygon-RNN · KITTI car instances | IOU: 74.22 -> 84.11 | Helped |
| Use Polygon-RNN automatic predictions rather than SharpMask | KITTI car annotation · KITTI car instances | IOU: 78.8 -> 74.22 | Hurt |
| Annotate image crops with object boxes and a disambiguating dot instead of free-viewing the full image | Human annotator · Cityscapes car instances | IoU: 69.5 -> 78.6 | Helped |
| Use Polygon-RNN corrections rather than GrabCut-assisted annotation | Semi-automatic Cityscapes annotation · 54 randomly chosen instances | IoU: 70.7 -> 79.7–85.8 | Helped |
Closest papers by meaning
- Efficient Interactive Annotation of Segmentation Datasets With Polygon-RNN++ (CVPR 2018 · similarity 0.92)
- Fast Interactive Object Annotation With Curve-GCN (CVPR 2019 · similarity 0.90)
- 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)
- Point-VOS: Pointing Up Video Object Segmentation (CVPR 2024 · similarity 0.84)
- Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty (NeurIPS 2026 · 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.