Comment by sorenjan

This is probably a stupid question, but what kind of image generation does this do? The architecture overview shows "input image", and I don't see anything about text to image. Is it super resolution? Does class-conditional mean that it takes a class like "car" or "face" and generate a new random image of that class?