A new technology, using artificial intelligence to significantly improve video quality, has been developed in Russia by specialists from Novosibirsk State University (NSU).
Video processing improves dozens of times after using neural networks. AI algorithms analyze image resolution, eliminate noise and artifacts in the video. The application will function on remote servers. It works through the Gradio web interface. Machine learning models are written in Python.
A one-minute video with a resolution of up to 480p can be processed in just 40 seconds, improving its quality to Full HD. While manual processing of each frame can take more than an hour.
Mikhail Solomin, one of the developers, said that generative adversarial neural networks are used for video processing. There are no analogues of this technology on the Russian market.
Read more on the topic:
Neural networks in Russia have been trained to recognize dangerous objects
MTUCI scientists will teach artificial intelligence to recognize damaged bar codes
Sber presented the concept of the Kandinsky Video 1.1 neural network
