Everypixel: A Portfolio Review via Neural Network

April 10, 2017

By Greg Scoblete

A neural network is a computing platform that’s been designed to mimic the architecture of the human brain. A new startup called Everypixel is leveraging a neural network to pass judgment on millions of stock photos in an effort to make finding quality stock photos easy.

The algorithm is still in beta, but you can test it on your own images here. As you can see, my humble beach snapshot has apparently just a 6 percent chance of being “awesome.”

According to Everypixel, the goal is to aggregate millions of stock images into a central repository and then create a search engine that ranks them by how beautiful they are. It also automatically tags images based on their contents. To date, the company is working on a database that includes images from 30 stock photo services.

Everypixel is hardly the first company to attempt to train an algorithm to make a subjectively human judgment on photos. EyeEm has a mature version of this technology, EyeEm Vision, helping to organize and surface images in its photo community. Other stock agencies, like Pond5, have implemented versions of this technology as well to facilitate automatic image tagging and to group similar images together.

Does Everypixel’s algorithm make the right call? Try it on your images and let us know in the comments.

Hat tip: DP Review