Robot finds Wally thanks to machine learning

Spread the love

Hobbyists have built a robot that looks at the pages of the books Where’s Wally? the boy with the red and white striped sweater. They trained the robot using Google’s AutoML Vision service.

The fastest time to find Wally, or Waldo, as the boy is called in the US, is 4.45 seconds, which seems to have room for improvement. The There’s Waldo system is therefore still a prototype, the makers emphasize.

The system is built on the basis of a Raspberry Pi, which controls the uArm Metal robotic arm using pyArm Python software. The arm takes pictures of the pages of the Where’s Wally? book and registers all the faces. All faces are sent by the system to Google’s AutoML Vision service, which makes the comparison with a model trained on Wally’s head.

When There’s Waldo is at least 95 percent sure there is a match, the arm moves to that position to point out Wally. Google AutoML allows users to train machine learning models, even if they have little or no prior knowledge. There’s Waldo is a project by creative agency RedPepper.

You might also like