Unsupervised-Attention-guided-Image-to-Image-Translation icon indicating copy to clipboard operation
Unsupervised-Attention-guided-Image-to-Image-Translation copied to clipboard

Change input image size

Open prash-p opened this issue 6 years ago • 3 comments

Changing IMAGE_HEIGHT and IMAGE_WIDTH in model.py to anything other than 256 x 256 gives rise to errors. How can the model be changed to adapt to any input image size (not necessarily square)?

prash-p avatar Dec 01 '18 19:12 prash-p

@prash-p have you experimented with images of varying input size? Would like to know what type of results you've been able to achieve

nickkimer avatar Jan 16 '19 15:01 nickkimer

When I tested in December, no other image size would work except 256 x 256, even when I changed the IMAGE_HEIGHT and IMAGE_WIDTH parameters. Would appreciate if @AlamiMejjati could direct me towards adjusting the parameters correctly for variable image size input

prash-p avatar Jan 16 '19 17:01 prash-p

@nickkimer went back to this repo and realised that you also need to change line 30 in main.py: self._size_before_crop to your desired size. It has to be the same size or larger than IMAGE_HEIGHT and IMAGE_WIDTH from models.py I'm experimenting to see what the results look like.

prash-p avatar Jan 30 '19 19:01 prash-p