Is it possible take a training set of one million 120 x 120 pixel tiles, feed these tiles into a machine learning algorithm. And then make it synthesize images that look like the original training set with a certain, adjustable level of entropy?

What ML techniques should I consider?


1 Answer 1


Yes, there are plenty ways to do this, what you basically want is to interpolate between the two.

This is current state-of-the-art for solving problems like yours: https://github.com/carpedm20/BEGAN-tensorflow

Also keep an eye out for my question here: 2D Image-based Avatar Animation With Expressions / Image-To-Image Translation

If I end up using the ML approach, you could very well also use it with tiles instead of faces and be able to adjust the likeness. I will link to my project in my question.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.