Researchers at Tencent have developed the GFP-GAN algorithm to repair corrupted images and low-resolution portraits. PetaPixel writes about it.
The technology combines information from two artificial intelligence models to fill in the missing details of photos in seconds, while maintaining high accuracy and quality.
Conventional image restoration methods require careful tuning of an existing AI model by measuring the differences between generated and real images, the scientists say. This often leads to substandard results, they added.
The new approach uses a pre-trained version of Nvidia’s StyleGAN-2 algorithm to inform the group’s own model at multiple stages of the imaging process. This preserves the “identity” of the people in the photograph, with special emphasis on facial features such as eyes and mouths.
The GFP-GAN demo is free. The researchers also published the source code of the model, allowing any developer to implement the recovery technology in their projects.
However, the creators of GFP-GAN warned that the algorithm could distort some of the details of faces and produce a lower resolution result.
Recall that in June, Adobe introduced a tool for restoring old photos in Photoshop.
In October 2021, Israeli startup D-ID developed a tool to “make” people in portraits talk.
In April, Deep Nostalgia taught Live Photos how to dance and blow kisses.
Subscribe to ForkLog news on Telegram: ForkLog AI – all the news from the AI world!
Found a mistake in the text? Select it and press CTRL+ENTER