Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Physics-Guided Deep Learning for Linear Mueller Matrix Acquisition
Yuan Liang, Shaoli Liu, Jiachun Huang
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 residual matrix refinement after pBRDF initialization (ablation) | PIPNet-pBRDF · Synthetic Mueller dataset | Forward Stokes L2: 0.4065 -> 0.0214 | Improved |
| add residual matrix refinement after pBRDF initialization (ablation) | PIPNet-pBRDF · Real Mueller dataset | Forward Stokes L2: 0.6450 -> 0.0926 | Improved |
| add pBRDF prior initialization to direct Stokes-to-matrix regression (ablation) | Direct Stokes->MM · Synthetic Mueller dataset | Forward Stokes L2: 0.0433 -> 0.0214 | Improved |
| add pBRDF prior initialization to direct Stokes-to-matrix regression (ablation) | Direct Stokes->MM · Real Mueller dataset | Forward Stokes L2: 0.1615 -> 0.0926 | Improved |
| use sequential pBRDF initialization instead of joint pBRDF/matrix multi-task (ablation) | Joint pBRDF/MM multi-task · Synthetic Mueller dataset | Forward Stokes L2: 0.0457 -> 0.0214 | Improved |
| use sequential pBRDF initialization instead of joint pBRDF/matrix multi-task (ablation) | Joint pBRDF/MM multi-task · Real Mueller dataset | Forward Stokes L2: 0.1965 -> 0.0926 | Improved |
| joint multi-task instead of direct Stokes-to-matrix (ablation) | Direct Stokes->MM · Synthetic Mueller dataset | Forward Stokes L2: 0.0433 -> 0.0457 | Hurt |
| our method vs best baseline Hwang et al. | Hwang et al. · Real Mueller dataset | Matrix PSNR: 28.27 -> 40.42 | Improved |
| recovered Mueller as prior for normal refinement instead of Stokes vector (ablation) | Initial Optimized normal · Normal refinement (synthetic) | Mean angular error: 7.706 -> 7.357 | Improved |
| use recovered Mueller instead of model-only Mueller prior (ablation) | Normal refinement · Normal refinement (synthetic) | Mean angular error: 8.316 -> 7.357 | Improved |
| use recovered Mueller instead of DoLP+AoLP features (ablation) | Normal refinement · Normal refinement (synthetic) | Mean angular error: 7.653 -> 7.357 | Improved |
| add polarizance P from recovered matrix instead of DoLP (ablation) | Material classification SVM · Material classification | Balanced accuracy: .7565 -> .8359 | Improved |
Closest papers by meaning
- PolarScale: A Physics-Grounded Benchmark for Radiometrically Consistent RGB-to-Stokes Estimation (NeurIPS 2026 · similarity 0.86)
- Glossy Object Reconstruction with Cost-effective Polarized Acquisition (CVPR 2025 · similarity 0.85)
- High-Fidelity Polarimetric Implicit 3D Reconstruction with View-Dependent Physical Representation (AAAI 2025 · similarity 0.84)
- PolarMatte: Fully Computational Ground-Truth-Quality Alpha Matte Extraction for Images and Video using Polarized Screen Matting (CVPR 2024 · similarity 0.84)
- OpticalNet: An Optical Imaging Dataset and Benchmark Beyond the Diffraction Limit (CVPR 2025 · similarity 0.84)
- Event Ellipsometer: Event-based Mueller-Matrix Video Imaging (CVPR 2025 · 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.
