Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Towards Robust Probabilistic Modeling on SO(3) via Rotation Laplace Distribution
Yingda Yin, Jiangran Lyu, Yang Wang, Haoran Liu, He Wang, Baoquan Chen
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 |
|---|---|---|---|
| replace Murphy et al. Implicit-PDF with rotation Laplace distribution | pose regressor · ModelNet10-SO3 (one model, all categories) | Acc@3: 0.294 -> 0.445 | Improved |
| replace Murphy et al. with rotation Laplace distribution | pose regressor · ModelNet10-SO3 (one model, all categories) | median error: 20.3 -> 13.0 | Improved |
| replace Murphy et al. with rotation Laplace distribution | pose regressor · Pascal3D+ (one model, PascalVOC_val) | Acc@30: 0.838 -> 0.874 | Improved |
| use rotation Laplace instead of quaternion Laplace (ablation) | Laplace-type distribution · ModelNet10-SO3 (one model, all categories) | Acc@3: 0.242 -> 0.445 | Improved |
| use rotation Laplace instead of RPMG-9D | per-category regressor · ModelNet10-SO3 chair | mean error: 11.9 -> 9.7 | Improved |
| use rotation Laplace instead of matrix Fisher with 30% outliers injected (ablation) | regressor · ModelNet10-SO3 with 30% injected outliers | Acc@30: 0.467 -> 0.700 | Improved |
| add Gaussian-like small label noise up to 30 degrees (fixed) to training labels (ablation) | rotation Laplace · ModelNet10-SO3 noise | Acc@15: 0.741 -> 0.618 | Hurt |
| increase normalization grids from 72 to 4608 (ablation) | rotation Laplace · ModelNet10-SO3 toilet | Acc@5: 0.76 -> 0.82 | Improved |
| vary clipping parameter eps from 1e-4 to 1e-12 (ablation) | rotation Laplace · ModelNet10-SO3 | Acc@3: 0.450 -> 0.441 | No clear effect |
| replace 6D rotation with rotation Laplace in GDR-Net | GDR-Net · YCB-video | ADD(-S): 49.1 -> 57.0 | Improved |
| replace 6D rotation with rotation Laplace in GDR-Net | GDR-Net · LINEMOD | ADD(-S): 93.7 -> 94.2 | Improved |
| replace matrix Fisher with rotation Laplace in FisherMatch | semi-supervised regression · ModelNet10-SO3 sofa 5% labels | mean error: 31.09 -> 26.12 | Improved |
Closest papers by meaning
- A Laplace-inspired Distribution on SO(3) for Probabilistic Rotation Estimation (ICLR 2023 · similarity 0.97)
- FisherMatch: Semi-Supervised Rotation Regression via Entropy-Based Filtering (CVPR 2022 · similarity 0.86)
- Normalizing Flows on the Product Space of SO(3) Manifolds for Probabilistic Human Pose Modeling (CVPR 2024 · similarity 0.84)
- Unified Framework for Diffusion Generative Models in SO(3): Applications in Computer Vision and Astrophysics (AAAI 2024 · similarity 0.83)
- Variational Multi-scale Representation for Estimating Uncertainty in 3D Gaussian Splatting (NeurIPS 2024 · similarity 0.82)
- Learning with 3D rotations, a hitchhiker's guide to SO(3) (ICML 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.
