Research map / 3D vision / Depth and stereo
Differentiable Programming for Hyperspectral Unmixing using a Physics-based Dispersion Model
John Janiczek, Parth Thaker, Gautam Dasarathy, Christopher S. Edwards, Philip Christensen, Suren Jayasuriya
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 |
|---|---|---|---|
| use dispersion-model analysis-by-synthesis instead of NCM | NCM · Feely (lab IR, 90 test samples) | abundance MSE: 0.119 -> 0.052 | Improved |
| train deep models on synthetic data with limited real data | CNN-1D · Feely (lab IR, 90 test samples) | abundance MSE: 0.205 -> 0.052 vs analysis-by-synthesis | Hurt |
| add generative dispersion model to CNN output (Inverse Rendering vs CNN-1D Modified) | CNN-1D Modified · Gulfport (VNIR aerial) | abundance MSE: 0.297 -> 0.272 | Improved |
| add generative dispersion model to CNN output (Inverse Rendering vs CNN-1D Modified) | CNN-1D Modified · Gulfport synthetic mixtures | abundance MSE: 0.148 -> 0.059 | Improved |
| add generative dispersion model to CNN output (Inverse Rendering vs CNN-1D Modified) | CNN-1D Modified · Feely (lab IR, 90 test samples) | abundance MSE: 0.205 -> 0.188 (both worse than analysis-by-synthesis 0.052) | Hurt |
| use analysis-by-synthesis (no training data) instead of FCLS | FCLS · Gulfport (VNIR aerial) | abundance MSE: 0.75 -> 0.45 | Improved |
| use BCM statistical variability model instead of analysis-by-synthesis | Analysis-by-synthesis · Gulfport synthetic mixtures | abundance MSE: 0.147 -> 0.136 | No clear effect |
| use analysis-by-synthesis on Mars TES data instead of FCLS | FCLS · Mars TES data | RMS reconstruction error: 0.0043 -> 0.0038 | Improved |
| use alternating minimization instead of joint optimization of abundances and dispersion parameters | joint optimization · Feely | speed and convergence | Improved |
| add dispersion-model analysis-by-synthesis runtime cost versus FCLS | FCLS · Unmixing speed (averaged over 90 mixtures) | time per operation: 10ms -> 10.2s | Hurt |
| add noise to input, methods trained without noise | CNN and Inverse Rendering · Feely noise test (30 samples) | abundance MSE under noise | Hurt |
Closest papers by meaning
- Solving Spectrum Unmixing as a Multi-Task Bayesian Inverse Problem with Latent Factors for Endmember Variability (AAAI 2024 · similarity 0.89)
- Unmixing Diffusion for Self-Supervised Hyperspectral Image Denoising (CVPR 2024 · similarity 0.87)
- Unmixing Before Fusion: A Generalized Paradigm for Multi-Source-based Hyperspectral Image Synthesis (CVPR 2024 · similarity 0.86)
- LESSViT: Robust Hyperspectral Representation Learning under Spectral Configuration Shift (NeurIPS 2026 · similarity 0.83)
- HIR-Diff: Unsupervised Hyperspectral Image Restoration Via Improved Diffusion Models (CVPR 2024 · similarity 0.83)
- FRN: Fractal-Based Recursive Spectral Reconstruction Network (NeurIPS 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.
