Research map / 3D vision / Face and shape
Learning to Reconstruct 3D Manhattan Wireframes From a Single Image
Yichao Zhou, Haozhi Qi, Yuexiang Zhai, Qi Sun, Zhili Chen, Li-Yi Wei, Yi Ma
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 offset regression for junctions (l2 loss) (ablation) | Joint wireframe network · SU3 validation (3D wireframe) | APC: 65.4 -> 72.8 | Improved |
| add offset regression for junctions (l2 loss) (ablation) | Joint wireframe network · SU3 validation (3D wireframe) | APT: 57.1 -> 60.1 | Improved |
| use l2 instead of l1 loss for offset regression (ablation) | Joint wireframe network · SU3 validation (3D wireframe) | APC: 69.3 -> 72.8 | Improved |
| use l2 instead of l1 loss for offset regression (ablation) | Joint wireframe network · SU3 validation (3D wireframe) | APT: 55.8 -> 60.1 | Improved |
| add line (edge) supervision jointly with junctions (ablation) | Joint wireframe network · SU3 validation (3D wireframe) | APC: 72.8 -> 74.3 | Improved |
| add line (edge) supervision jointly with junctions (ablation) | Joint wireframe network · SU3 validation (3D wireframe) | APT: 60.1 -> 61.0 | Improved |
| add junction supervision jointly with lines (ablation) | Joint wireframe network · SU3 validation (3D wireframe) | line heatmap IoU: 73.3 -> 74.2 | Improved |
| add ordinal loss to depth supervision (ablation) | Joint wireframe network · SU3 validation (3D wireframe) | SILog: 3.59 -> 4.14 | Hurt |
| add joint junction and line training to depth estimation (ablation) | Joint wireframe network · SU3 validation (3D wireframe) | SILog: 3.59 -> 3.04 | Improved |
| train junctions and lines jointly instead of separately (ablation) | Wireframe network · Wireframe dataset (2D) | AP: 67.8 -> 71.0 | Improved |
| use proposed network instead of prior parser with separate training | 2D wireframe detection · Wireframe dataset (2D) | AP: 67.5 -> 67.8 | Improved |
| use neural vanishing point regression instead of LSD/J-linkage | Vanishing point detection · SU3 vanishing point detection | failure rate: 20.0% -> 2.3% | Mixed |
Closest papers by meaning
- Learning to Parse Wireframes in Images of Man-Made Environments (CVPR 2018 · similarity 0.93)
- PC2WF: 3D Wireframe Reconstruction from Raw Point Clouds (ICLR 2021 · similarity 0.88)
- NEAT: Distilling 3D Wireframes from Neural Attraction Fields (CVPR 2024 · similarity 0.88)
- Neural Wireframe Renderer: Learning Wireframe to Image Translations (ECCV 2020 · similarity 0.87)
- EdgeDiff: Edge-aware Diffusion Network for Building Reconstruction from Point Clouds (CVPR 2025 · similarity 0.84)
- Wired Perspectives: Multi-View Wire Art Embraces Generative AI (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.
