-4

I have a collection of 3500 images and each image is apparted of 12 merged figures that look like this:

enter image description here

I am searching for an unsupervised ML algorithm that will help me identify possible clusters out of these images. Any suggestions?

Jokerp
  • 213
  • 1
  • 9
  • 1
    Downvoting as the question is related to methodology. Kindly look at the help centre on what sort of questions should be asked. Now, as for your question, you should use any backbone such as `vgg16` to `efficientnet` and get the features that you can now pass on to KMeans or any other algorithm. – Abhishek Prajapat Aug 10 '21 at 13:11

1 Answers1

1

You could for example use K-means or DBSCAN (use the pixels as coordinates).

Then you could highlight your clusters by changing the pixel colors.

kodkirurg
  • 156
  • 8