Skip to content

CRISM-ML-unsupervised

Latest
Compare
Choose a tag to compare
@beatricebs beatricebs released this 10 Sep 16:25

First version of a new toolkit to perform unsupervised machine learning (clustering) on CRISM near-infrared hyperspectral data on Mars. The toolkit will:

  • Read and visualise CRISM data;
  • Perform data preprocessing;
  • Reduce the dimensionality of the data through projection-based (PCA) and, additionally, manifold-based (UMAP) techniques;
  • Perform clustering on the data through the available methods (k-Means and Gaussian Mixture Models);
  • Extract and visualise the mean spectra of each cluster;
  • Evaluate cluster quality with the Silhouette score.

The tool has been developed with the support of Dr. Mario D'Amore (@kidpixo)