Is it possible to classify images using an unsupervised learning algorithm ???

Question

I want to classify images into two groups for gender identification (face images) using any unsupervised algorithm

Answer 1

It's technically possible but it's classifications would still need to be moderated for accuracy. Most image recognition systems have a database of images they are trained on, and the algorithm then matches key points in the image to the database and picks the statistically most likely option (e.g. a blue flat thing with light reflecting off it might be a window or some water, but if it has straight/smooth edges it is more likely to be a window).

It ultimately depends what you want the algorithm to do; if you want to see what happens then go for it and tell us about it, if its for a specific purpose you may be better off giving it a training set first.

Answer 2

That would require more time/samples than supervised learning.

Answer 3

I’d say so, yes.

google and apple have done it

Answer 4

Labeling unlabeled data, maybe. It might work better to help identify anomalous patterns