# Recognizing hand-written archaeological signs

I have 75 images of handwritten signs from which I extracted 7 Hu moments and solidity features. How can I find similarities among them to train a classifier and predict the value? I thought SVM was a good choice, but I don't have a target vector (what do I put? I do not know differences in signs that I can say there are, say, three labels, such as "circle", "triangle" or "square").

Is cluster analysis more appropriate? If yes, What is best method? I am using the scikit-learn module in python.

http://scikit-learn.org/stable/modules/svm.html

http://scikit-learn.org/stable/modules/clustering.html

-
– postgres Feb 6 '13 at 15:32
Can you show us the full set of samples? – Emre Feb 6 '13 at 21:02
– postgres Feb 6 '13 at 22:53

It is a bit hard to understand what you are trying to do. What are these signs? The one you posted looks like a wheel. Are there meaningful categories that you can name? If so, then this is a supervised learning (classification) problem, and you should use a classification algorithm such as SVM.

If there are no clear labels, but you want to group together similar-looking signs, then this is an unsupervised learning problem, and you should use a clustering algorithm.

From your description it sounds more like a clustering problem.

-
– postgres Feb 6 '13 at 22:53
I'm not sure if it is better affinity propagation mean shift or dbscan with my image set, and after i found clusters there is any function that tell me most similar images inside the cluster? – postgres Feb 7 '13 at 16:01