Clustering results are discarded - new release about pix-plot

[duhaime]

Thanks for reaching out @AGrosserHH ! As you noticed, we've moved to a python module setup, but it's easy enough to find and modify the module to get this information.

First things first, one should find where the pixplot module code is. To do that, you should be able to type python and hit enter in a terminal. That will drop you into the Python interpreter (you'll know you're in the interpreter if you see >>> as your terminal prompt).

Once you're in the interpreter, try typing:

import pixplot
print(pixplot.__file__)

That should tell you the path like:

.../pixplot-{version}/pixplot/__init__.py

If you navigate to the folder that contains that init.py file, you should see another file pixplot.py; that's the file we'll want to change to display the info you discuss above.

Once you open up that file, you'll see a function get_centroids(). In that function, there's a variable z with an attributelabels_ that contains the cluster assignments of each input image in order (so the ith value in z.labels_ is the cluster id for the ith image in kwargs['image_paths']). To get the number of images in each cluster and the names of the images in each cluster, one can therefore add (somewhere after z is fully populated):

cluster_map = defaultdict(list)
  for idx, i in enumerate(z.labels_):
    cluster_map[i].append(kwargs['image_paths'][idx])
  for i in cluster_map:
    print(' * cluster {} has {} images: {}'.format(i, len(cluster_map[i]), cluster_map[i]), '\n')

I hope that helps! I'm happy to follow up if you try to add this and get stuck...

[duhaime]

@AGrosserHH I'm going to close this issue for now, but please feel free to follow up if you have any troubles!