One of the great doubts of a photographer or anyone who takes photos even with his mobile is if that snapshot is good enough to share. The stock photo portal Everypixel has created its own neural network that tells you whether or not it is of good quality.
A neural network is a computer system that allows you to interpret elements in a similar way to as a person would do it. In this case, it is based on algorithms trained to decide if a photo is good. This Everypixel system has been trained with more than 950,000 photographs with positive and negative votes.
The neural network interprets the elements by which it believes that a photo has been voted negative, learning to detect the same elements in other photos. For example, if it detects a face and is not well positioned, the focus is confusing or not where it should be focused, the color of the photo or the elements in it.
Based on these thousands of examples organized by publishers, photographers and associates of Everypixel, it can calculate the quality level of the photo in percentage. Of course, we are talking about art, therefore the quality of a photo is more in the eye that sees it, that of a programmed system.
For example, highly prized photos such as Tetons and the Snake River by photographer Ansel Adams only gets a 32.3% chance of liking it. Why? Consider the tone of the photo, the colors, the focus, the brightness or elements like the framing. In this example, the photo is considered a work of art.
These neural networks have much to appreciate the real talent. Another thing there are the photos that you can take while you learn or of stock, which is where this company wants to use it, to help us to choose better photos.