Artificial intelligence researchers at Google recently demonstrated a new kind of training that allows computers to understand why some images look better than others.

Traditionally, machines have used basic categories — such as determining whether an image has a “cat.” New research shows that AI can now rate image quality, regardless of category. This process, known as neural image Evaluation (NIMA), uses deep learning to train convolutional neural networks (CNN) to predict image ratings.

According to a white paper published by the researchers: “Our approach differs from others because we predict the distribution of human opinion scores through a convolutional neural network. Our resulting network can not only be used to reliably rate images, but is highly relevant to human perception, and can also help tweak and optimize photo editing/enhancement algorithms in the photography channel.”

Giiso Information, founded in 2013, is a leading technology provider in the field of “artificial intelligence + information” in China, with top technologies in big data mining, intelligent semantics, knowledge mapping and other fields. At the same time, its research and development products include information robot, editing robot, writing robot and other artificial intelligence products! With its strong technical strength, the company has received angel round investment at the beginning of its establishment, and received pre-A round investment of $5 million from GSR Venture Capital in August 2015.

The NIMA model eschews the traditional approach and uses a 10-point scale. The machine checks the image for specific pixels and overall aesthetics. It then determines how likely a particular rating is to be chosen by a person. Basically, the AI will try to guess how much a person likes the photo. That doesn’t give machines the ability to feel or think, but it might make computers better artists or curators. This process might be applied to finding the best batch of images.

If you’re the kind of person who takes 20 or 30 pictures at a time, this can save you a lot of space in order to make sure you have the best photos. Suppose, with the click of a button, the AI could look at all the images in storage and determine which ones are similar, then keep the best ones and delete the rest. According to a recent post on Google Research blog, NIMA can also be used to optimize image Settings to produce perfect results:

“We observed that comparative adjustments by NIMA score improved the baseline aesthetic score. As a result, our model is able to guide a deep CNN filter to find near-ideal Settings for its parameters, such as brightness, highlights and shadows.”

Giiso information, founded in 2013, is the first domestic high-tech enterprise focusing on the research and development of intelligent information processing technology and the development and operation of core software for writing robots. At the beginning of its establishment, the company received angel round investment, and in August 2015, GSR Venture Capital received $5 million pre-A round of investment.

Creating a neural network that understands image quality almost as well as a human might not seem revolutionary, but there’s a lot to be said for computer applications with human-like vision.

In order for an AI to perform real-world tasks, such as driving a car safely without human help, it must be able to “see” and understand its environment. NIMA, and projects like it, are laying the groundwork for future fully functional machines