Welcome to the 6-Band Hyperspectral Classification & Conformal Uncertainty Quantification Workspace.
This repository provides reproducible pipelines across 11 architectures and 6 loss function trials.
| Module | Description | Coverage Level |
|---|---|---|
| baseline/ | Standard CNN Backbones (AlexNet, GFNet, ViT_UNet) | 95.0% |
| credit/ | Confidence Regularized Evidential Deep Inference | 95.0% |
| ensemble/ | Deep Ensembles & CreDE Uncertainty | 95.2% |
| multicp/ | Multi-Head Conformal Prediction | 95.1% |
| sacp/ | Self-Adaptive Conformal Prediction | 95.0% |
| spatialgcn/ | Spatial-Spectral Graph Convolutional Network | 95.4% |
| mambahsi/ | State-Space Mamba Hyperspectral Vision | 96.3% |
| dofa/ | Dynamic Optical Foundation Model & HieraViT | 95.2% |
| georsclip/ | Geospatial Remote Sensing CLIP | 95.0% |
| dapm/ | Diffusion-Augmented Probabilistic Model | 95.0% |
| trials/ | 6 Loss Functions (CB-Focal, CDL, Conformal-Reg, EDL, EDL-v2, Focal) | Comprehensive |