Old photo restoration using neural networks
Now you won’t surprise anyone with filters that improve the quality of photos. But the restoration of old portraits still leaves much to be desired. Older photos tend to be too blurry, so normal image sharpening methods won't work on them.
The TantcentARC team has developed a new algorithm to improve old photo portraits based on a combination of two methods Generative Facial Prior - Generative Adversarial Network (GFP-GAN).
The first method tries to isolate any digital noise and improve face areas on the photo, while the second method tries to reject unsuccessful options generated by the first one.
According to the researchers, this approach allows to achieve a higher quality of blind face restoration compared to previous methods and algorithms.
Anyone can evaluate the work of scientists and try to restore old snaps at https://huggingface.co/spaces/akhaliq/GFPGAN
More details about the GFP-GAN method can be found at https://github.com/TencentARC/GFPGAN