№3 (46), 2018. INFORMATION TECHNOLOGY, COMPUTER SCIENCE AND CONTROL

Zelensky A.A., Franz V.A.

Generative adversarial neural networks for video feature extraction in task of visual quality assessment

The paper explores possible ways of training video quality metrics on small datasets. We propose a new approach for video feature extraction based on generative adversarial networks. We show that combination of generative adversarial network framework with other regularization techniques enables training of visual quality metrics with high level of conformity to subjectively perceived quality level on datasets currently available and typically used for visual quality evaluation.

Keywords: deep learning, blind visual quality metrics, objective video quality assessment, adversarial learning.

[ Back to contents ]
[ Get the article ]