search

mousecpn / Joint-Bilateral-Learning

An unofficial implementation of Joint Bilateral Learning for Real-time Universal photorealistic Style Transfer
MIT License
67 stars 23 forks source link

test.py question #5

Open mzprose opened 4 years ago

mzprose commented 4 years ago

If my test picture is a picture of any resolution size, such as 1k, 2k, can I omit this step?

transform = transforms.Compose([ transforms.Resize((512, 512), Image.BICUBIC), transforms.ToTensor() ])

if not ,what can I do

mousecpn commented 4 years ago

If my test picture is a picture of any resolution size, such as 1k, 2k, can I omit this step?

transform = transforms.Compose([ transforms.Resize((512, 512), Image.BICUBIC), transforms.ToTensor() ])

if not ,what can I do

It seems that the model can not be applied to other resolutions.

mzprose commented 4 years ago

However, the author stated in the original text that it can be inferred at a higher resolution. The resize (512*512) limits this theory. Where is the problem, can you analyze the reason for me?

mousecpn commented 4 years ago

However, the author stated in the original text that it can be inferred at a higher resolution. The resize (512*512) limits this theory. Where is the problem, can you analyze the reason for me?

I figure it out. Low resolution must be (256,256), but the high resolution cont img can be any scale.

mzprose commented 4 years ago

ok,thanks for your reply.