Research map / 3D vision / 3D generation
Semi-supervised Semantic Mapping Through Label Propagation with Semantic Texture Meshes
Radu Alexandru Rosu, Jan Quenzel, Sven Behnke
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add label propagation retraining (ablation) | Single frame · NYUv2 | mean IoU: 0.302 -> 0.322 | Improved |
| add label propagation retraining (ablation) | SemanticFusion · NYUv2 | mean IoU: 0.324 -> 0.343 | Improved |
| add label propagation retraining (ablation) | Ours · NYUv2 | mean IoU: 0.363 -> 0.372 | Improved |
| use textured semantic mesh instead of SemanticFusion surfel map | SemanticFusion · NYUv2 | mean IoU: 0.324 -> 0.363 | Improved |
| retrain on worst (highest inconsistency) frames instead of best frames (ablation) | label propagation · NYUv2 | IoU | Improved |
| run more label propagation iterations (ablation) | label propagation · NYUv2 | IoU | No clear effect |
| add label propagation on courtyard (ablation) | single-frame prediction · Courtyard | mean IoU: 7% | Improved |
| add label propagation on courtyard fused semantics (ablation) | mesh semantics · Courtyard | mean IoU: 3% | Improved |
| retrain RefineNet on pseudo ground truth (ablation) | RefineNet · Mapillary | IoU: 0.4546 -> 0.4487 | Hurt |
| add linear distance fall-off 0-100 m instead of no weighting (ablation) | semantic integration · Courtyard | IoU: 0.294 -> 0.3289 | Improved |
| restrict d_max to 30 m (ablation) | semantic integration · Courtyard | IoU: 0.3289 -> 0.309 | Hurt |
| change d_min to 30 m (ablation) | semantic integration · Courtyard | IoU: 0.3289 -> 0.3259 | No clear effect |
Closest papers by meaning
- SceneCode: Monocular Dense Semantic Reconstruction Using Learned Encoded Scene Representations (CVPR 2019 · similarity 0.90)
- MAP-ADAPT: Real-Time Quality-Adaptive Semantic 3D Maps (ECCV 2024 · similarity 0.89)
- Think While You Map: Asynchronous Vision-Language Agents for Incremental 3D Scene Graphs (ECCV 2026 · similarity 0.87)
- Semi-supervised 3D Semantic Scene Completion with 2D Vision Foundation Model Guidance (AAAI 2025 · similarity 0.87)
- SNI-SLAM: Semantic Neural Implicit SLAM (CVPR 2024 · similarity 0.87)
- SUM Parts: Benchmarking Part-Level Semantic Segmentation of Urban Meshes (CVPR 2025 · similarity 0.86)
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.
