Questions about test the multi images on a pre-trained model
Opened this issue · 1 comments
Hi, I want to know how well this kind of "single-image" GAN model shows on multi images. I have seen in #6 that these models maybe work well on the images which are similar to the trained image and could not get the desired result for different images. Therefore, is there some other research about how to solve this problem, or you will do further work about it?
Thank you for your excellent work!
Thank you for the kind words and for your question.
This methods learns the distribution of patches of an image. It would not work for many images as is. A different approach is needed in order to have this generally trained. There are several approaches we considered. There are many single-feed-forward retageting papers, but it is not quite the same as what we do.