sunshineatnoon/PytorchWCT

Using I_5 as the input of the VGG RELU_4_1?

Closed this issue · 1 comments

Deleted user commented

Hi,

First of all, I appreciate your work and code - your implementation is clean, efficient and easy to follow :-)

Nevertheless, I found that the RELU_4_1 takes the original content image as input:

PytorchWCT/WCT.py

Line 59 in 2a2b4e4

cF4 = wct.e4(contentImg)

I believe it should take Im5 (the intermediate result generated from d5 ) as input, no?

sunshineatnoon commented

Sure you can change that to RELU_5_1, I just thought RELU_4_1 gives a better result.