EyeEm’s Software Knows What’s Pretty

A person doesn’t “decide” a photo is beautiful—the reaction is emotional, instinctive. So when Appu Shaji began trying to duplicate the process with code four years ago, building a program to rank photos by their beauty, he failed. A computer imaging researcher at EPFL, an engineering college in Lausanne, Switzerland, Shaji designed his software to judge photos based on composition, color saturation, and perspective, but couldn’t teach it to identify the best-looking ones. “With hard rules, I wasn’t able to solve the problem,” he says.

So Shaji rounded up Berlin’s best photographers and artists and asked them to judge the same images. By tweaking his code based on their input, he says, he slowly made the software good enough to match the taste of a person instead of a machine. Last year, Berlin software startup EyeEm bought Shaji’s company, Sight.io, for an undisclosed amount and made him the head of research and development.