Technology can help you with various tasks and demands today and one of its most recent advantages is the possibility of restore old photos. Developed by Chinese researchers, a tool created with the help of artificial intelligence can recover photos damaged by the action of time. This is an excellent mechanism to keep some memories alive and materialized!
Read the article and learn more about the tool.
Read more: Learn how to send photos without losing quality
How it works
Everyone has old photos at home, many of them damaged, smudged, or low resolution. As the physical recovery of the images is difficult and expensive, many people ended up losing some memories. However, it is now possible to use an online tool to make the photo look brand new. Developed by the company Tencent, in China, the platform called GFP-GAN, Generative Facial Prior-Generative Adversarial Network, is capable of correcting flaws in images without altering the original photo.
The software seeks to remove the fold marks from the photos, highlight barely visible or confusing details and highlight the colors of the images, including transforming black and white photos into color images. The method, based on the artificial intelligence technology StyleGAN-2, is based on a probability system.
If other tools add elements to the image that make them artificial, the GFP-GAN platform seeks to maintain the naturalness of the photos. To do so, it does not add new information to the photos and only seeks to clean up the scenery. If it is necessary to fill in some space in the photo, the elements are collected through a database. Even to improve aspects of the face, such as mouth, nose and eyes.
The application is valid both for old photographs in low resolution and for digital images. As camera photos can be damaged or corrupted at the time of capture, the tool also successfully recovers these images.
To use the application, no payment is required. The demo version is available to everyone at GFP-GAN page or by free download at GitHub site. The application is open source software, that is, it can be downloaded and used by anyone. Thus, the intention of the creators is to allow the algorithm to be used for custom tasks that can be integrated into new software.
The researchers estimate that a more advanced model of the application will be released in a few weeks. Among the improvements listed are the improvement of facial recognition, identification of elements with low resolution in the photo and better filling of spaces.