this notebook allows to classify images according to their similarity without choosing the number of clusters.
To make it work, you have to put the directory of images in the same path as the notebook and to rename it 'dataset'. Then you simply launch the notebook.