Vision / ML2024
Featured Project
Gesture Controlled Rock Paper Scissors
This game uses MediaPipe landmark detection to gather information about the player's hand's position. These landmarks are fed into a custom model trained using Scikit Learn while the computer generates random actions to compete with.
Problem
A gesture game needs more than decent model output. It needs a full realtime loop from hand detection to feature extraction to a response that feels immediate enough to play against.
Solution
I used MediaPipe to extract hand landmarks, trained a custom Scikit Learn model for gesture classification, and connected the predictions to a rock-paper-scissors game loop against a computer opponent.