You can find AI that generates new images, but what if you want to restore an old family photograph? You might have a free option.
GFP-GAN (Generative Facial Prior-Generative Adversarial Network), a free tool recently developed by Tencent researchers, can restore damaged and low-resolution portraits, according to Louis Bouchard and PetaPixel.
MORE FROM RAVZGADGET: Amazon Sues Operators Of 10,000 Facebook Groups Over Fake Reviews
The technology combines information from two AI models to fill in missing details in a photo with realistic detail in a matter of seconds, while maintaining high accuracy and quality.
Conventional methods fine-tune an existing AI model to restore images by comparing artificial and real photos. According to the scientists, this frequently results in low-quality results.
The team’s own model is informed at multiple stages during the image generation process by a pre-trained version of an existing model (NVIDIA’s StyleGAN-2).
The technique aims to preserve people’s “identities” in photographs, with a particular emphasis on facial features such as eyes and mouths.
GFP-GAN demo is available for free. The creators have also released their source code, allowing anyone to use the restoration technology in their own projects.
MORE FROM RAVZGADGET: Google Marks Play Store’s 10th Birthday With A New Logo
This project is still constrained by the limitations of current artificial intelligence. While it is surprisingly accurate, it is only guessing about missing content. The researchers cautioned that you may notice a “slight change in identity” and a lower resolution than you prefer.
Please do not rely on this to print a poster-sized photo of your grandparents. Nonetheless, the work here is promising — it suggests a future in which images can be easily saved that would otherwise be lost to the ravages of time.
Reading your article helped me a lot, but I still had some doubts at the time, could I ask you for advice? Thanks.