GFP-GAN: free AI tool for old photos

GFP-GAN is a free, open source, AI-based tool that will restore your old family photos.

gfp gan

Most there are too many circulating photo recovery tools that work with artificial intelligence. But if you want to fix an old family photo (and more), try Tencent's GFP-GAN program.

This is a free and open source photo recovery tool that takes only a few seconds to fix an old photo.

GFP-GAN (Generative Facial Prior-Generative Adversarial Network) is almost unique. Unlike most AI photo recovery tools, it aims to fix old photos without hiding their subject's identity.

This means it focuses on cleaning up noise, correcting wrinkles, bringing out detail and enhancing color with little guesswork.

As you can see in the images above, GFP-GAN doesn't add much to the photos. It just cleans things up and fills in some gaps when needed. Subjects do not lose their facial hair or mutate into something alien. This is partly due to StyleGAN-2 by NVIDIA, which the Tencent researchers use alongside their own model.

Artificial intelligence recovery tools are never perfect, and GFP-GAN has some failures too. Tencent researchers warn that the restored photos may not have sufficient resolution and that in some cases, family members may look a bit strange

However you can try GFP-GAN in your browser or download it source code on Github.

iGuRu.gr The Best Technology Site in Greecefgns

Subscribe to Blog by Email

Subscribe to this blog and receive notifications of new posts by email.

GFP-GAN,artificial,intelligence,photo,picture

Written by Dimitris

Dimitris hates on Mondays .....

Leave a reply

Your email address is not published. Required fields are mentioned with *

Your message will not be published if:
1. Contains insulting, defamatory, racist, offensive or inappropriate comments.
2. Causes harm to minors.
3. It interferes with the privacy and individual and social rights of other users.
4. Advertises products or services or websites.
5. Contains personal information (address, phone, etc.).