PoseMan

What if a break from the keyboard could become a game?

PoseMan explores that idea through a webcam and a simple challenge: make shapes with your body and guess the letters. Your movement becomes the input, rather than another click or keypress.

The project combines pose estimation with a second neural network to recognise the poses. An animated character responds to the webcam feed, bringing the prediction back into the game.

It sits at an intersection I enjoy: an approachable interaction on the surface, with an interesting technical problem underneath.

JavaScript / ml5.js / TensorFlow / p5.js