Style transfer tutorial we need to clamp the values to [0,1]?

1

I have done style transfer and it’s variants on tensorflow, but never came across this problem. Can anyone explain why we need to clamp the values between [0,1], and why are the values ranging from - inf to inf.

Thank you

2

Well because that we don’t have any innate constraint on that so it is there to avoid the image go beyond range. In most situations, most pixels shouldn’t be problems.