menyifang / DCT-Net

Official implementation of "DCT-Net: Domain-Calibrated Translation for Portrait Stylization", SIGGRAPH 2022 (TOG); Multi-style cartoonization
Apache License 2.0
774 stars 76 forks source link

getting blurry and improper results in all the styles #77

Open dummyuser-123 opened 2 months ago

dummyuser-123 commented 2 months ago

Hello @menyifang,

First of all, thank you for this great work. It successfully transforms any face image into a cartoon style. While trying out different styles on various images, I noticed some issues in the outputs.

Input Image Styled Image
027 3d_cartoon_027
031 3d_cartoon_031
033 3d_cartoon_033
025 cartoon_025
031 cartoon_031

1) In images 1, 2, and 3, the corresponding outputs have a black line between the eye and the eyebrows in the 3D cartoon style. 2) In the cartoon style, the output image has issues in the facial part. It is not able to properly cartoonize some parts of the face. 3) Additionally, I am getting blurry images as output while trying any of the styles. I have tried with different image sizes, but I am not getting clear outputs.

It would be great if you could share some insights and experiences related to these problems and provide a solution or suggest if I am making any mistakes while running the code.