Louis Bouchard and PetaPixel have drawn attention to a free tool recently developed by Tencent researchers, GFP-GAN (Generative Facial Prior-Generative Adversarial Network), that can repair corrupted and low-resolution photos, according to Engadget.< /p>
The free technology combines information from two artificial intelligence models to fill in the missing details of a photo with realistic details in a few seconds while maintaining high accuracy and quality.
Anyone can try the GFP-GAN demo for free.
Conventional methods fine-tune the existing AI model for image reconstruction by measuring the differences between artificial and real photographs. According to scientists, this often leads to substandard results. The new approach uses a pre-trained version of an existing model (NVIDIA StyleGAN-2) to inform the group's own model at multiple stages of the imaging process. This technique aims to preserve the “identity” of the people in the photograph, with particular emphasis on facial features such as eyes and mouths.
You can try the GFP-GAN demo for free. The creators have also published their code, allowing anyone to implement the recovery technology in their projects.
This project is still bound by the limitations of current AI. While it's surprisingly accurate, it makes educated guesses about missing content. The researchers warned that you might see “a slight change in personality” and a lower resolution than you'd like. Don't rely on this to print a poster-sized photo of your grandparents. However, the work here is promising – it hints at a future where you can easily rescue images that would otherwise be lost to the ravages of time.
The system uses a pre-trained version of the existing NVIDIA StyleGAN neural network model -2. Its fine tuning was aimed at preserving the “identity” of the person in the original photo, with special emphasis on facial features such as eyes and mouth.