Using a live feed from a security camera, Olsen was able to collect a series of images of their dog pooping that they then used to train an AI model called DeepLabCut that’s able to recognize specific poses in animals. In this case, the pose Olsen’s dog makes when it’s relieving itself in the backyard. Every time the AI model detected a deposit was made, it highlighted the spot in the backyard with a red circle on the live feed from the security camera, making it easier to round up all the poop afterwards, but not as easy as it could be.

Pinpointing where all the red circles on the security camera live feed lined up on the backyard lawn was sometimes tricky, so Olsen further expanded the system with the addition of a bright green laser attached to a robotic arm. Once outside and on-camera, Olsen can trigger the robot by simply crossing their arms in an X. The system then calculates the nearest poop and projects a green dot on the ground in front of Olsen that physically guides them to its location. When another trained gesture is detected—Olsen crouching down to scoop the poop—the system moves on to the next pile, and then the next, slowly leading Olsen on what has to be the least fulfilling scavenger hunt imaginable.


It may seem like an over-indulgent use of technology, but any dog owner who’s accidentally discovered a missed pile of backyard poop by accidentally stepping on it would disagree. It’s ingenious, and not only useful for a backyard. Imagine every public park in the country should installing something like this, even just as a warning system to those strolling through.