Research map / Point clouds / Segmentation and detection
UnderOneFacade: Worldwide Facade Semantic Segmentation Benchmark Dataset
Yi Wang, Fan Wang, Prabin Gyawali, Ziyang Xu, Anna Klimkowska, Yixiong Jing, Wanru Yang, Filip Biljecki, Christoph Holst, Benjamin Busam, Brian Sheil, Olaf Wysocki
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 RGB and intensity features to xyz input (ablation) | PointNet++ · UK LoFG3 | mIoU: 22.1 -> 29.4 | Improved |
| add RGB and intensity features to xyz input (ablation) | PTv1 · UK LoFG3 | mIoU: 31.0 -> 39.1 | Improved |
| add RGB and intensity features to xyz input (ablation) | DGCNN · UK LoFG3 | mIoU: 28.3 -> 30.2 | Improved |
| add RGB and intensity features to xyz input (ablation) | DGCNN · UK LoFG3 | overall accuracy: 77.8 -> 76.5 | Hurt |
| add RGB and intensity features to xyz input (ablation) | PointNet++ · UK LoFG3 window F1 | F1: 13.7 -> 48.8 | Improved |
| add RGB and intensity features to xyz input (ablation) | PointNet++ · UK LoFG3 terrain F1 | F1: 95.4 -> 89.1 | Hurt |
| use graph-based DGCNN instead of transformer PTv1 | facade segmentation · UnderOneFacade LoFG3 overall | mIoU: 15.1 -> 33.4 | Improved |
| use DGCNN instead of PTv3 | facade segmentation · UnderOneFacade LoFG3 overall | mIoU: 27.3 -> 33.4 | Improved |
| train on Asia (Singapore) and test on Europe instead of reverse | PointNet++ · zero-shot LoFG3 | mIoU: 14.0 -> 23.6 | Improved |
| train on Asia and test on Europe instead of reverse | PTv3 · zero-shot LoFG2 | overall accuracy: 60.9 -> 53.4 | Hurt |
| test on Singapore instead of UK (cross-country shift) | PT · LoFG3 wall F1 | F1: 83.7 -> 43.8 | Hurt |
| test on Singapore instead of UK (cross-country shift) | PT · LoFG2 opening F1 | F1: 97.7 -> 6.8 | Hurt |
Closest papers by meaning
- Efficient 2D and 3D Facade Segmentation Using Auto-Context (TPAMI 2017 · similarity 0.87)
- 3D Semantic Parsing of Large-Scale Indoor Spaces (CVPR 2016 · similarity 0.87)
- Geometric Exploitation for Indoor Panoramic Semantic Segmentation (NeurIPS 2024 · similarity 0.86)
- HouseLayout3D: A Benchmark and Training-free Baseline for 3D Layout Estimation in the Wild (NeurIPS 2025 · similarity 0.86)
- XMask3D: Cross-modal Mask Reasoning for Open Vocabulary 3D Semantic Segmentation (NeurIPS 2024 · similarity 0.84)
- Large Spatial Model: End-to-end Unposed Images to Semantic 3D (NeurIPS 2024 · 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.
