Tencent has developed an artificial intelligence system to restore old
Software

Tencent has developed an artificial intelligence system to restore old snapshots

Tencent engineers have created an artificial intelligence (AI) system called GFP-GAN (Generative Facial Prior-Generative Adversarial Network) that aims to improve quality and repair damage to old photos. Based on data from two AI models, it fills in the missing parts of photos in seconds, offering fairly high accuracy and quality.

    Image source: youtube.com

Image source: youtube.com

Typically, such projects are a series of fine-tuning of existing AI models that are trained on the differences between artificial and real images. According to Tencent, this approach often produces inferior results. Alternatively, they trained the NVIDIA StyleGAN-2 model through several standard steps in the image generation process, which allowed them to retain a person’s “identity” in the original photo and focus on the key features of the face, such as the eyes and mouth.

The authors of the project decided to make the results available to everyone. You can try the algorithm either Demo implementationor use code system if you want to use it in your project. The project is far from perfect, but even now its assumptions about the missing image fragments seem pretty accurate. The researchers warned that some changes in real people’s facial features are still possible and the system cannot yet work at very high resolutions.

About the author

Robbie Elmers

Robbie Elmers is a staff writer for Tech News Space, covering software, applications and services.

Add Comment

Click here to post a comment