Research map / 3D vision / 3D generation
MAP-ADAPT: Real-Time Quality-Adaptive Semantic 3D Maps
Jianhao Zheng, Daniel Barath, Marc Pollefeys, Iro Armeni
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 geometric-complexity-based splitting (SG) on top of semantic-only splitting (S) (ablation) | MAP-ADAPT · HSSD middle regions (@4cm) | Completion error (cm): 2.89 -> 2.67 | Improved |
| add geometric-complexity-based splitting (SG) on top of S (ablation) | MAP-ADAPT · ScanNet coarse regions (@8cm) | Completion error (cm): 5.01 -> 4.27 | Improved |
| add geometric-complexity splitting (SG) | MAP-ADAPT · HSSD coarse regions (@8cm) | Geometric error (cm): 4.53 vs 4.05 for Multi-TSDFs | No clear effect |
| add geometric complexity (SG vs S) in map size and runtime (ablation) | MAP-ADAPT · HSSD | Map size (MB): 265.21 -> 469.85 | Hurt |
| use adaptive map instead of fixed 1cm Voxblox | Voxblox 1cm · HSSD | Map size (MB): 1225.30 -> 265.21 | Improved |
| add adaptive raycasting (ablation) | MAP-ADAPT-SG · HSSD fine regions (GT input) | Completion error (cm): 0.40 -> 0.29 | Improved |
| add adaptive raycasting (ablation) | MAP-ADAPT-SG · HSSD coarse regions (GT input) | Completion error (cm): 0.86 vs 0.85 | No clear effect |
| split neighboring voxels when a voxel is refined (ablation) | MAP-ADAPT-SG · HSSD fine regions (GT input) | Geometric error (cm): 0.67 -> 0.37 | Improved |
| adaptive resolution vs fixed fine resolution in fine regions | MAP-ADAPT-S · HSSD fine regions (@1cm) | Completion error (cm): 2.49 vs 2.54 | No clear effect |
Closest papers by meaning
- Semi-supervised Semantic Mapping Through Label Propagation with Semantic Texture Meshes (IJCV 2019 · similarity 0.89)
- Think While You Map: Asynchronous Vision-Language Agents for Incremental 3D Scene Graphs (ECCV 2026 · similarity 0.87)
- Understanding while Exploring: Semantics-driven Active Mapping (NeurIPS 2025 · similarity 0.86)
- SNI-SLAM: Semantic Neural Implicit SLAM (CVPR 2024 · similarity 0.85)
- Change-Robust Online Topological Memory for Long-Term Relocalization and Semantic Navigation (NeurIPS 2026 · similarity 0.85)
- CymbaDiff: Structured Spatial Diffusion for Sketch-based 3D Semantic Urban Scene Generation (NeurIPS 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.
