You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Thank you for integrating the ReStyle encoder into JoJoGAN. I’ve been fine-tuning JoJoGAN using your code for a cartoon style, but I haven’t been getting the desired results. The ReStyle encoder should ideally produce better results compared to e4e, but in my case, the e4e encoder is yielding better outcomes than the ReStyle encoder.
My Training Details:
Alpha = 1
iterations = 300
no of style images = 1
preserve color = False
Original Image
Styled Image
Styled Image:
As you can see from the results, the output seems to mimic the structure of the cartoon image rather than preserving the original image. Additionally, there are some unwanted red, green, and yellow shades in the output images. Could you please help me understand why I am getting these results and how I might improve them?
The text was updated successfully, but these errors were encountered:
Hello @007prateekd,
Thank you for integrating the ReStyle encoder into JoJoGAN. I’ve been fine-tuning JoJoGAN using your code for a cartoon style, but I haven’t been getting the desired results. The ReStyle encoder should ideally produce better results compared to e4e, but in my case, the e4e encoder is yielding better outcomes than the ReStyle encoder.
My Training Details:
Styled Image:
As you can see from the results, the output seems to mimic the structure of the cartoon image rather than preserving the original image. Additionally, there are some unwanted red, green, and yellow shades in the output images. Could you please help me understand why I am getting these results and how I might improve them?
The text was updated successfully, but these errors were encountered: