Restore Faces are more crisp and good then restored_imags

[rohaantahir]

Hi, I am using this great library using following settings.
python inference_gfpgan.py -i inputs/upload -o results -v 1.4 -s 1 --suffix enhanced --bg_upsampler none -s 2 -only_center_face --bg_tile 512 --bg_upsampler realesrgan

I am facing an issue that the resultant restored faces are very much identical to ground truth however, in restored images it is not showing perfect results.

The restored face image is following:

The restored whole image is following:

Can we have a similar kind of result for restored whole image as well ? If not do we have any solution to use restored faces? when we have a video the restored faces have bumps that's why couldn't use the restore faces.

Thanks

[Fifd]

[rohaantahir]

Any update @Fifd ??