View the demo video and complete source code on GitHub:
Marshmallow Shooter
An ML-Driven Human-Machine Interaction Study
The Marshmallow Shooter is a gamified autonomous dispenser designed to explore real-time facial tracking and precision ballistic delivery. By integrating MediaPipe computer vision with a bespoke dual-axis mechanical gantry, the system identifies a user’s facial coordinates and mouth state to "train" humans in the art of catching airborne snacks.
Core Functions:
Targeting: Real-time X/Y head tracking via Wekinator-trained servos.
Triggering: Mouth-open gesture recognition via MediaPipe Face Mesh.
Ballistics: High-velocity propulsion using a dual-motor friction drive.