A Free AI Tool Can Restore Old Photos Wonderfully
According to Louis Bouchard and PetaPixel, there is a free tool that can restore portraits damaged by time and having a low resolution. This tool was developed by researchers from Tencent and was named Generative Facial Prior-Generative Adversarial Networt, shortened to GFP-GAN.
Indeed, it is very irritating not to be able to repair, for example, an old family photo, but with GFP-GAN, it becomes possible. To do this, the technology used consists of merging information from two AI models to complete the missing details. The most interesting thing is that everything is done in just a few seconds.
How does the GFP-GAN work?
The improved version 1.3 of the GFP-GAN model tries to analyze the exact content of the photograph to understand it. She will then fill in what is missing and add pixels if necessary to the missing sections. PetaPixel says that ” the pretrained StyleGAN-2 model is then used to guide its own multi-scale generative model during encoding from image to latent code and thus to reconstruction “.
Facial details are greatly enhanced by AI through the use of additional measurements. AI focuses on important local features including a person’s eyes, mouth and nose. When this is done, the system starts comparing the real image to the image that was just restored to make sure it’s still the same person in the generated photo. The last step therefore aims to preserve the “identity” of all those who are in a photo.
The limits of the GFP-GAN
For information, anyone can try this fascinating restoration technology since the demo of GFP-GAN is completely free. Moreover, it can be integrated into the various projects of a person because the code of GFP-GAN has been made public by its creators.
However, it should be known that the current AI has its limits, which means that this technology also has some. According to the researchers, there could be a “slight identity change” as well as a lower resolution than what you had in mind. It is therefore not recommended to rely on such a tool to print a large format version of an old photograph of great-grandparents. But this does not remain a great technological advance.
source
