Suppose you have a data set composed of n images as training examples. You run clustering on each image ( initializing 3 clusters per image) and learn the centers. Is it ok to then take the cluster centers themselves as features for a supervised learning algorithm and thus have a vocabulary for each image that way or is it inconsistent ? Are there are other more consistent measures that can be used ?
closed as not a real question by Jeffε, Tsuyoshi Ito, Lev Reyzin♦, David Eppstein, Suresh Venkat Jan 30 '13 at 9:24
It's difficult to tell what is being asked here. This question is ambiguous, vague, incomplete, overly broad, or rhetorical and cannot be reasonably answered in its current form. For help clarifying this question so that it can be reopened, visit the help center. If this question can be reworded to fit the rules in the help center, please edit the question.
I'd say you can. "Coordinates of point of importance inside image" qualifies as a feature so you can use it in supervised learning. If clustering is a way to discover that coordinate you can use that.
A practical example: say you want to use "x-coordinate of left eye" as a feature in a classification problem. It does not matter how you get the object's values for that feature, as long as the values are accurate. If you have an algorithm (e.g. a clustering algorithm) which can compute the values you can use it.