Deep-Photo-Enhancer
Deep-Photo-Enhancer copied to clipboard
Any re-implementation in pytorch?
For some unknown reason, the code runs very slow in my device and is hard to debug for me. Only the initialization takes almost 10 minutes.
Has anyone be able to re-implement the unpairwise results using pytorch? I have tried for several days using the fiveK dataset but only obtained the highest psnr of 18.0 and the visual results look unnatural like follows:
first row: input, second: model output, last: expert C
Somehow, this model is not taking as much time to load in my case. Is it possible to share the exact images which you are using, so that I can test them with this current implementation and tell you how the output looks like using TF code?
You should select the InputAsShotZeroed as the input data.
Somehow, this model is not taking as much time to load in my case. Is it possible to share the exact images which you are using, so that I can test them with this current implementation and tell you how the output looks like using TF code?
Hi, I'm trying to use Pytorch to re-impleament the code, but for now I only got the highest psnr of 10.8
Hi,
When I run the TF.py in the "[Online_Demo_Models]_Deep-Photo-Enhancer/LPGAN_exp_G3_736" file, I face this error:
Shapes (2, 1, 32, 32, 128) and () are incompatible
in line: netG_test_output1 = model(netG, test_df.input1, False, netG_act_o, is_first=True)
Would you mind helping me with it? Thank you in advance.