Research map / Image generation / Image restoration
Synthesis and Completion of Facades from Satellite Imagery
Xiaowei Zhang, Christopher May, Daniel Aliaga
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 grammar-parameter optimization (ablation) | facade synthesis pipeline · 61 test facades | F-score: 0.597 -> 0.815 | Improved |
| add grammar-parameter optimization (ablation) | facade synthesis pipeline · 61 test facades | blob accuracy: 0.810 -> 0.923 | Improved |
| add grammar-parameter optimization (ablation) | facade synthesis pipeline · 61 test facades | pixel accuracy: 0.725 -> 0.880 | Improved |
| use procedural grammar approach instead of Nishida et al. | facade subdivision · 61 facades | MAE floors: 0.770 -> 0.246 | Improved |
| use procedural grammar approach instead of Nishida et al. | facade subdivision · 61 facades | MRE windows per floor: 12.1% -> 3.9% | Improved |
| use full pipeline instead of Pix2Pix 96 segmentation | facade parsing · 61 facades | F-score: 0.758 -> 0.815 | Improved |
| use full pipeline instead of Pix2Pix 96 segmentation | facade parsing recall · 61 facades | recall: 0.868 -> 0.834 | Hurt |
| use full pipeline instead of DeepFill inpainting then Pix2Pix 96 | facade parsing · 22 occluded facades | F-score: 0.677 -> 0.783 | Improved |
| add DeepFill inpainting before Pix2Pix 96 | facade parsing · 22 occluded facades | F-score: 0.728 -> 0.677 | Mixed |
| choose Pix2Pix 96 over other segmentation models (ablation) | segmentation · 61 facades | pixel accuracy: 0.843665 -> 0.870966 | Mixed |
Closest papers by meaning
- Visual-to-Executable Procedural Reconstruction of Buildings. (NeurIPS 2026 · similarity 0.87)
- Structured Outdoor Architecture Reconstruction by Exploration and Classification (ICCV 2021 · similarity 0.86)
- Sat2City: 3D City Generation from A Single Satellite Image with Cascaded Latent Diffusion (ICCV 2025 · similarity 0.85)
- Regularity-Driven Facade Matching Between Aerial and Street Views (CVPR 2016 · similarity 0.85)
- 3D Building Reconstruction from Monocular Remote Sensing Images with Multi-level Supervisions (CVPR 2024 · similarity 0.83)
- SG-BEV: Satellite-Guided BEV Fusion for Cross-View Semantic Segmentation (CVPR 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.
