Two friends, Manon Bickert and Marine Paquet have exhibited, in the Insitut de Physique du Globe de Paris, a collection of pictures of minerals through the lens of a microscope. I have seen that machine learning algorithms can be applied to a lot of different images to create clusters and classify the pixels.
I have found two way to separate them so far:
- the
k-meansalgorithm - the
agglomerative-clusteringalgorithm
These two approaches come from the scikit-learn package, so make sure you have it installed before starting.
I've been using the following example but these codes should work with any kind of pictures:
The k-means will cluster minerals of similar color wherever they are in the figure (here for n_cluster = 6):
The agglomerative-clustering will cluster minerals of similar color only if they are close to each other (here for n_cluster = 10):
