Google introduced the Neural Image Assessment (NIMA) system, which uses a neural network to evaluate photos, but not only their technical specifications, but and aesthetic beauty.
How it works
Artificial Intelligence in Google was trained on photos that experts have already appreciated. Experts assessed general view of the image, and put up points for each specific criterion. The system has analyzed photo ratings - and externally passed the exam art history.
What are the successes
FROM Using special algorithms, NIMA estimated more than 2,000 real images. Experts also looked at the images and evaluated them by scale from 1 to 10 points. AT resulting in ratings of AI more than 80% converged with opinion of critics.
AT Google says that NIMA will be useful in applications for working with photos. For example, she can quickly sort images by selecting the best. Also, the system is useful when editing a photo: the neural network will be able to adjust the parameters of exposure, brightness, contrast, saturation , etc.
Source: The Verge
Read in Russian: Google превратил ИИ в искусствоведа