Research map / Point clouds / Point cloud classification
PointAR: Efficient Lighting Estimation for Mobile Augmented Reality
Yiqin Zhao, Tian Guo
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 |
|---|---|---|---|
| learn from point cloud with PointConv instead of projected panorama with ResNet50 (ablation) | PointAR · Matterport3D/Neural Illumination test | SH coefficients l2 loss: 0.781 -> 0.633 | Improved |
| learn from point cloud with PointConv instead of projected panorama with ResNet50 (ablation) | PointAR · Matterport3D/Neural Illumination test | irradiance map l2 loss: 0.535 -> 0.433 | Improved |
| increase point cloud size from 512 to 1280 points (ablation) | PointAR · Matterport3D/Neural Illumination test | SH coefficients l2 loss: 0.668 -> 0.633 | Improved |
| increase point cloud size from 512 to 1280 points (ablation) | PointAR · Matterport3D/Neural Illumination test | irradiance map l2 loss: 0.479 -> 0.433 | Improved |
| increase point cloud size from 1024 to 1280 points (ablation) | PointAR · Matterport3D/Neural Illumination test | irradiance map l2 loss: 0.441 -> 0.433 | Improved |
| increase point cloud size from 768 to 1280 points (ablation) | PointAR · Matterport3D/Neural Illumination test | SH coefficients l2 loss: 0.660 -> 0.633 | Improved |
| increase point cloud size from 512 to 1280 points (ablation) | PointAR · Model complexity | MACs (M): 320 -> 790 | Hurt |
| increase point cloud size from 512 to 1280 points (ablation) | PointAR · Model complexity | parameters (M): 1.42 -> 1.42 | No clear effect |
| use PointAR point cloud pipeline instead of Garon et al. image network | PointAR vs Garon et al. · Matterport3D/Neural Illumination test | irradiance map l2 loss: 0.63 -> 0.433 | Improved |
| use PointAR point cloud pipeline instead of Garon et al. image network | PointAR vs Garon et al. · Matterport3D/Neural Illumination test | SH coefficients l2 loss: 1.10 -> 0.633 | Improved |
| use PointAR instead of Song et al. multi-stage network | PointAR vs Song et al. · Matterport3D/Neural Illumination test | irradiance map l2 loss: 0.619 -> 0.433 | Improved |
| use PointAR instead of ResNet50 backbone in terms of cost | PointAR · Model complexity | parameters (M): 25.56 -> 1.42 | Improved |
Closest papers by meaning
- Fast Spatially-Varying Indoor Lighting Estimation (CVPR 2019 · similarity 0.90)
- DeepLight: Learning Illumination for Unconstrained Mobile Mixed Reality (CVPR 2019 · similarity 0.86)
- HDR Environment Map Estimation for Real-Time Augmented Reality (CVPR 2021 · similarity 0.86)
- Deep Parametric Indoor Lighting Estimation (ICCV 2019 · similarity 0.86)
- LightOctree: Lightweight 3D Spatially-Coherent Indoor Lighting Estimation (CVPR 2024 · similarity 0.85)
- LiDAR-Net: A Real-scanned 3D Point Cloud Dataset for Indoor Scenes (CVPR 2024 · similarity 0.83)
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.
