Thanks for your paper and offering this code. I'm building on your code and modifying it for a different domain that has lower frequency textures than the plant images you used. I was wondering, how did the quality of your generated visible and/or thermal images turn out? I don't recall (and sorry if you reported it and I didn't see it), in your paper, if you reported FID scores or other metrics about the quality of the generated images. Thanks!