---
language:
- en
license: mit
library_name: keras
tags:
- remote-sensing
- multispectral
- hyperspectral
- uncertainty-quantification
- conformal-prediction
metrics:
- accuracy
- kappa
- coverage
---

# 6-Band Hyperspectral Classification Model Registry

This repository hosts deep neural network checkpoints trained on the 6-band hyperspectral benchmark dataset (330x307x6, 7 classes) for conformal uncertainty quantification and high-precision remote sensing classification.

## Benchmark Architectures
- **AlexNet CNN Backbone**: 6-channel convolutional feature extractor
- **GFNet (Global Filter Network)**: Frequency-domain 2D FFT global spatial interaction network
- **ViT_UNet**: Hybrid Vision Transformer and U-Net encoder-decoder architecture

## Checkpoints Included
- `baseline/models/AlexNet_CNN_best.keras`
- `baseline/models/GFNet_best.keras`
- `baseline/models/ViT_UNet_best.keras`
- `baseline/models/AlexNet_CNN_final.keras`
- `baseline/models/GFNet_final.keras`
- `baseline/models/ViT_UNet_final.keras`

## Authors & Attribution
- **Principal Author**: Danesh Selwal (`selwaldanesh@gmail.com`)
- **Co-authored by**: Claude (`claude@anthropic.com`)