CRISM-ML-unsupervised

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)