Dodging the Trump
Punching the Trump

StumpTrump

Can you stump the trump?

Built with openCV and python, stump Trump in an experiment of webcam based control in video games. In Stump Trump, one interacts with the game by punching towards the icons on screen, keeping the player active and providing a slightly more immersive experience.

Webcam based tracking poses some interesting problems to solve, especially when considering what assumptions you can make about the user's environment, and what external factors you have to account for. For example, without proper lighting, feature detection algorithms may not even work, and when they do, they are very slow. Since StumpTrump uses feature detection to track the player's face in the dodging level, we had to find a way to preserve the functionality of the feature detection, but without the slow or unreliable processes. Our solution was to only detect the user's head on startup (or when the key 'r' is pressed) and then track the motion of the head without attempting to detect it further. This does reduce accuracy to a certain extent, but it also speeds up the process, and allows the player to sit in comfortable positions or move around slightly without causing the detector to fail.

This technology can be implemented in scenarios other than gaming. Physiotherapy, martial arts tutorials, and as a form of computer input in general are some examples.