Yuchen Cai Zachary Berry Jeremy Kofman Heidi Luah

This unique game is designed with a setup featuring a rail with 4 mobile apertures and 2 webcams situated at each end, capturing the hand movements of the two players positioned on either side. These mobile apertures, when closed, act as wings that obscure the players' faces and their hand gestures, adding an intriguing challenge to deciphering the signs they're making. To progress, each player must verbally describe the hand sign they're forming, and only when the webcams simultaneously recognize both players making the correct sign, will one of the openings fully unveil, revealing the players' faces and adding a layer of excitement and connection to the game.

We're utilizing Handpose OSC to handle the input data. This system receives video feed from the webcam and employs a pre-trained model that accurately estimates the 3-dimensional (XYZ) position of 21 key points on the hand, which translates to 63 input data points for Wekinator.

In our setup, Wekinator features two outputs. We've trained a regressive neural network, specifically designed for two outputs. These outputs correspond to two distinct hand gestures we hope to be able to recognize.

SYSTEM DIAGRAM

CIRCUIT DIAGRAM

APERTURE MECHANICAL DIAGRAMS

FINAL SETUP

FULL PDF HERE