Research map / Detection and segmentation / CNN features
Fast Interactive Object Annotation With Curve-GCN
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 sequential polygon prediction with simultaneous graph-based spline prediction | Object annotation on Cityscapes · Cityscapes | mean IoU: 71.38 -> 73.70 | Helped |
| Add differentiable accuracy loss after matching-loss training (ablation) | Spline-GCN · Cityscapes | mean IoU: 72.09 -> 73.70 | Helped |
| Add iterative inference (ablation) | Spline-GCN · Cityscapes | mean IoU: 68.55 -> 70.00 | Helped |
| Add boundary prediction branch to the CNN (ablation) | Spline-GCN · Cityscapes | mean IoU: 70.00 -> 72.09 | Helped |
| Use a spline rather than polygon representation (ablation) | Curve-GCN · Cityscapes | mean IoU: 67.79 -> 68.55 | Helped |
| Use spline control points with differentiable accuracy optimization rather than PSP-DeepLab | Automatic object annotation · Cityscapes | mean IoU: 73.66 -> 73.70 | No clear effect |
| Add four extreme-point clicks as model input (ablation) | Spline-GCN · Cityscapes | mean IoU: 73.70 -> 79.88 | Helped |
| Add one boundary correction click to box-per-component input (ablation) | Spline-GCN-MBOX · Cityscapes | mean IoU: 77.62 -> 80.19 | Helped |
| Use only 10% of target-domain training data to fine-tune (ablation) | Cityscapes-trained Spline-GCN · KITTI, ADE20K, Aerial Rooftop, and Cardiac MR | mean IoU: 84.09 -> 84.81; 72.94 -> 77.35; 68.33 -> 78.21; 78.54 -> 91.33 | Helped |
| Use local interactive corrections rather than recurrently updating subsequent vertices | Interactive Curve-GCN · Cityscapes interactive annotation | click efficiency | Helped |
| Use local interactive corrections | Interactive Curve-GCN · Cityscapes interactive annotation | click count | Mixed |
Closest papers by meaning
- Efficient Interactive Annotation of Segmentation Datasets With Polygon-RNN++ (CVPR 2018 · similarity 0.93)
- Annotating Object Instances With a Polygon-RNN (CVPR 2017 · similarity 0.90)
- SplineCNN: Fast Geometric Deep Learning With Continuous B-Spline Kernels (CVPR 2018 · similarity 0.85)
- Large-Scale Interactive Object Segmentation With Human Annotators (CVPR 2019 · similarity 0.84)
- FeaStNet: Feature-Steered Graph Convolutions for 3D Shape Analysis (CVPR 2018 · similarity 0.84)
- Box2Poly: Memory-Efficient Polygon Prediction of Arbitrarily Shaped and Rotated Text (AAAI 2024 · similarity 0.82)
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.