This AI-Powered Camera Only Shoots ‘Award-Winning’ Photos

The Trophy Camera is an experimental camera powered by artificial intelligence that can only shoot images that it deems to be “award-winning.”

Using machine learning, media artist Dries Depoorter and PhD student/photographer Max Pinckers trained the camera to recognise common patterns in World Press Photo winners from 1955 until today. When a photograph is taken, the camera gives it a rating based on how closely it matches the attributes of prior winners. Photos that get a 90% grade or higher are automatically posted to a website called in all their glory.

The camera was built to remove all control from the user –– the main components are a Raspberry Pi computer, a red button switch, and OLED screen that simply reads out text including the photo’s grade in lieu of a viewfinder.

In an interview with Fast Co. Design, Pinckers reveals that the camera was designed as commentary on the march towards automation in photography, where cameras are perfect machines that create what he sees as redundant imagery.

“Press photography appears to be becoming a self-referential medium dominated by tropes, archetypes, and pop-culture references. What implications does this have on how we learn about the world through the images we are being shown?”, Pinckers told Fast Co. Design.

“By making this camera, we try to implicitly comment on the current status of photojournalism–which seems to be becoming more questionable in today’s visual landscape–along with the incredibly fast development of computer vision and the relevance of artificial intelligence in our time.”

The actual photos that the conceptual Trophy Camera take are hardly award-winning, consisting mainly of blurry images of people viewing the camera at an exhibition at FOMU, Antwerp.

Pinckers and Depoorter hope that through their work, people will put more thought in to the media that they absorb on a daily basis.

(via Fast Co. Design via DPReview)

Image credits: Photographs by Max Pinckers and Dries Depoorter