Research map / 3D vision / Face and shape
ArchSym: Detecting 3D-Grounded Architectural Symmetries in the Wild
Hanyu Chen, Ruojin Cai, Steve Marschner, Noah Snavely
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 |
|---|---|---|---|
| predict signed distance maps relative to predicted geometry instead of direct plane regression | ArchSym · ArchSym test (19 scenes) | Geodesic distance (deg): 5.06 -> 3.71 | Improved |
| predict signed distance maps instead of direct plane regression | ArchSym · ArchSym test (19 scenes) | Dense symmetry error: 0.18 -> 0.13 | Improved |
| use ArchSym model instead of finetuned Reflect3D | ArchSym · ArchSym test (19 scenes) | Geodesic distance (deg): 10.46 -> 3.71 | Improved |
| use ArchSym model instead of finetuned Reflect3D | ArchSym · ArchSym test (19 scenes) | F@15deg: 0.55 -> 0.84 | Improved |
| add FiLM instance-specific conditioning (ablation) | ArchSym · ArchSym test (19 scenes) | Geodesic distance (deg): 6.72 -> 3.71 | Improved |
| add FiLM instance-specific conditioning (ablation) | ArchSym · ArchSym test (19 scenes) | Dense symmetry error: 0.20 -> 0.13 | Improved |
| use pseudo-GT signed distance from predicted geometry instead of GT geometry supervision (ablation) | ArchSym · ArchSym test (19 scenes) | Geodesic distance (deg): 6.99 -> 3.71 | Improved |
| use pseudo-GT signed distance from predicted geometry instead of GT geometry supervision (ablation) | ArchSym · ArchSym test (19 scenes) | Dense symmetry error: 0.19 -> 0.13 | Improved |
| parameterize plane by normal and point on plane instead of offset | Direct baseline · ArchSym test (19 scenes) | training stability / accuracy | Improved |
| use frozen VGGT backbone features for direct regression vs object-trained Reflect3D | Direct baseline · ArchSym test (19 scenes) | Geodesic distance (deg): 10.46 -> 5.06 | Improved |
Closest papers by meaning
- Symmetry Strikes Back: From Single-Image Symmetry Detection to 3D Generation (CVPR 2025 · similarity 0.91)
- Learning to Detect 3D Symmetry From Single-View RGB-D Images With Weak Supervision (TPAMI 2022 · similarity 0.90)
- Beyond Planar Symmetry: Modeling Human Perception of Reflection and Rotation Symmetries in the Wild (ICCV 2017 · similarity 0.89)
- E3Sym: Leveraging E(3) Invariance for Unsupervised 3D Planar Reflective Symmetry Detection (ICCV 2023 · similarity 0.89)
- Leveraging 3D Geometric Priors in 2D Rotation Symmetry Detection (CVPR 2025 · similarity 0.88)
- GIQ: Benchmarking 3D Geometric Reasoning of Vision Foundation Models with Simulated and Real Polyhedra (ICLR 2026 · 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.
