Google is Developing Artificial Intelligence Capable of Critiquing Photos

Everyone is a critic… and soon, even Google will have the ability to add its two cents about your photography. Google’s Artificial Intelligence is developing NIMA: Neural Image Assessment, which is a program that can be trained to view and rate photographs based not just on their technical quality (such as resolution) but on other criteria that some might call “artistic merit.”

Using statistical models and similar images that get high ratings, the program scans an image and compares it with how other images fall on a ratings scale of 1-10 for each of several criteria; this is a method used in other forms of machine learning. The more photos the program “sees,” the more data it has to base its ratings of new photos.

The program takes all the different criteria, as well as comparisons against a huge database for Aesthetic Visual Analysis. Each photo in this database has been scored by human beings — an average of 200 people — in response to photography contests. Once the program goes through all these photos and their human ratings, the theory goes, it will then be able to look at new photos and rate them based on similar criteria.

In tests so far, the AI’s ability to score images based on aesthetics came scarily close to the humans’.

The program could even take it a step further. NIMA scores can be used to automatically enhance photos to boost their ratings, although the samples Google provides look like they simply had their clarity and dynamic range enhanced.

So far, it’s still in the research stage; the white paper, which you can read here, was published in late December. Will there be a Google Photo Critic in our future? Only time will tell.