Encoder reconstruction yields very bad results

[javierbenitezmarin]

Hello,
I have been trying to reproduce the harmonization task, and still haven't managed to get the results shown as examples.

I use the provided face encoder "ghfeat-encoder-ffhq-256.pkl" and the "image.list" file provided in the examples folder.
When performing the feature extraction, the script takes about 20 minutes to finish, and the resulting reconstructed images look like this:

I have performed several tests, using the bedroom encoder too, and the results are always the same:

:

When installing the environment, I changed the decorator library from "decorator==5.0.0" to "decorator" (decorator=5.0.0 yields an error when trying to install).

My distro is Ubuntu 20.04.3 LTS.

[xyIsHere]

Hi,
I got the same results as you. Did you solve this problem?

[javierbenitezmarin]

Hey!
Nope, I assumed it was a hardware problem of the graphics card, because the same exact code with the same exact models worked fine on another system..
But yeah, I gave up trying to make it work as I felt like I was wasting time. Sorry for not being able to help :(

[xyIsHere]

Hey! Nope, I assumed it was a hardware problem of the graphics card, because the same exact code with the same exact models worked fine on another system.. But yeah, I gave up trying to make it work as I felt like I was wasting time. Sorry for not being able to help :(

Thank you so much! It is very helpful information for me. I think you are right. I also tried to run the training code. It seems that my driver and cuda version is too high. It is really hard to run the tf1.x code with my current environment.