Research map / 3D vision / Depth and stereo
VPLNet: Deep Single View Normal Estimation With Vanishing Points and Lines
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 Manhattan line maps and vanishing-point-guided Manhattan labels to normal estimation | UPerNet normal estimator · ScanNet | mean angle error: 14.23 -> 13.76 | Helped |
| Add Manhattan line maps and vanishing-point-guided Manhattan labels to normal estimation | ScanNet-trained normal estimator · NYUD-v2 | mean angle error: 21.60 -> 17.98 | Helped |
| Add Manhattan line maps and vanishing-point-guided Manhattan labels to normal estimation | ScanNet-trained normal estimator · Replica | mean angle error: 18.23 -> 17.08 | Helped |
| Remove the Manhattan line map input (ablation) | vp-line model · ScanNet | mean angle error: 13.76 -> 13.93 | Hurt |
| Remove the Manhattan label-map prediction (ablation) | vp-line model · ScanNet | mean angle error: 13.76 -> 14.14 | Hurt |
| Use only Manhattan labels to produce the combined normal output instead of the final fused output (ablation) | vp-line model · ScanNet | mean angle error: 13.76 -> 14.08 | Hurt |
| Use the raw normal prediction instead of the final fused output (ablation) | vp-line model · ScanNet | mean angle error: 13.76 -> 13.92 | Hurt |
| Initialize with semantic-segmentation pretrained weights (ablation) | baseline normal estimator · ScanNet | mean angle error: 16.90 -> 14.23 | Helped |
| Initialize with semantic-segmentation pretrained weights (ablation) | vp-line model · ScanNet | mean angle error: 15.33 -> 13.76 | Helped |
| Use jointly predicted depth and normals instead of depth-only prediction (ablation) | depth network · ScanNet | RMSE: 0.342 -> 0.280 | Helped |
| Provide a pretrained predicted normal map as an input (ablation) | depth network · ScanNet | RMSE: 0.342 -> 0.267 | Helped |
Closest papers by meaning
- Unsupervised Learning of Geometry From Videos With Edge-Aware Depth-Normal Consistency (AAAI 2018 · similarity 0.89)
- Deep Iterative Surface Normal Estimation (CVPR 2020 · similarity 0.89)
- GeoNet: Geometric Neural Network for Joint Depth and Surface Normal Estimation (CVPR 2018 · similarity 0.88)
- Just Look at the Image: Viewpoint-Specific Surface Normal Prediction for Improved Multi-View Reconstruction (CVPR 2016 · similarity 0.88)
- Geometry and Learning Co-Supported Normal Estimation for Unstructured Point Cloud (CVPR 2020 · similarity 0.88)
- PMNI: Pose-free Multi-view Normal Integration for Reflective and Textureless Surface Reconstruction (CVPR 2025 · similarity 0.87)
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.