Research map / Point clouds / Segmentation and detection
Semantic Instance Annotation of Street Scenes by 3D to 2D Label Transfer
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 |
|---|---|---|---|
| Use joint 3D-to-2D label transfer instead of 3D point projection followed by graph cuts | Semantic label transfer · Suburban street scenes, 120 frames | JI: 80.5 -> 89.2 | Helped |
| Use joint 3D-to-2D label transfer instead of the strongest 2D baseline | Semantic label transfer · Suburban street scenes, 120 frames | Acc: 87.4 -> 94.2 | Helped |
| Add 3D point terms to LA+PW+CO (ablation) | Semantic label transfer · Semantic segmentation, 160 images | JI: 85.2 -> 88.2 | Helped |
| Add 3D point terms to LA+PW+CO (ablation) | Instance label transfer · Instance segmentation, 160 images | JI: 78.4 -> 87.5 | Helped |
| Add 2D pairwise constraints to local appearance (ablation) | Semantic label transfer · Semantic segmentation, 160 images | JI: 82.1 -> 84.4 | Helped |
| Add 3D primitive constraints to LA+PW (ablation) | Semantic label transfer · Semantic segmentation, 160 images | JI: 84.4 -> 85.2 | Helped |
| Add 3D pairwise constraints (ablation) | Semantic label transfer · Semantic segmentation, 160 images | JI: 88.2 -> 88.8 | Helped |
| Add fold and curb potentials (ablation) | Semantic label transfer · Semantic segmentation, 160 images | JI: 88.8 -> 89.0 | Helped |
| Add fold and curb potentials (ablation) | Instance label transfer · Instance segmentation, 160 images | JI: 87.9 -> 88.1 | Helped |
| Select pixels by lowest predicted label entropy instead of labeling every pixel (ablation) | Full Model · Semantic segmentation, 160 images at 90% pixel density | JI: 89.0 -> 94.9 | Helped |
Closest papers by meaning
- Associatively Segmenting Instances and Semantics in Point Clouds (CVPR 2019 · similarity 0.88)
- Aerial Lifting: Neural Urban Semantic and Building Instance Lifting from Aerial Imagery (CVPR 2024 · similarity 0.87)
- SUM Parts: Benchmarking Part-Level Semantic Segmentation of Urban Meshes (CVPR 2025 · similarity 0.85)
- Seg2Box: 3D Object Detection by Point-Wise Semantics Supervision (AAAI 2025 · similarity 0.85)
- 3D StreetUnveiler with Semantic-aware 2DGS - a simple baseline (ICLR 2025 · similarity 0.85)
- LogoSP: Local-global Grouping of Superpoints for Unsupervised Semantic Segmentation of 3D Point Clouds (CVPR 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.