---
license: cc-by-4.0
task_categories:
- feature-extraction
pretty_name: Data-Source
size_categories:
- 1M<n<10M
tags:
- geospatial
- remote-sensing
- spatial-analysis
- benchmarking
---

This repository contains the dataset used in our work. The dataset has been prepared for thee following Indian statese : 

<table>
  <tr>
    <td>
      <ul>
       <li><strong>Sikkim</strong></li>
        <li><strong>Uttarakhand</strong></li>
        <li><strong>Karnataka</strong></li>
      </ul>
    </td>
    <td>
      <img src="https://cdn-uploads.huggingface.co/production/uploads/65feb1326f427c0e92295989/bO62R7_CPYnKMneOINyqD.png" width="300" height="400" alt="Indian map with marked states" />
    </td>
   </tr>
</table>

The final output is a multi-modal dataset containing terrain, spectral, and texture-based features extracted from satellite data, along with manually annotated landslide zones for training machine learning models.

# Structure of the ML Ready Dataset

Each ZIP archive corresponds to a single state and contains multi-modal data derived from remote sensing sources, including:

* Raw Sentinel-2 image tiles
* Digital Elevation Model (DEM) data
* GLCM-based texture features
* Spectral index feature maps
* Spatial subsets
* Manually annotated landslide regions