Machine Learning Maze

During Spring 2019, I conducted an independent study titled, "Machine Learning in Visual Studies." The goal of this course was to explore the applications of machine learning in creativity and art, interactive design, and generative models. For my final project, I investigated the use of machine learning in gaming. I created a physical maze that users could interact with using gestures that they determine. First, in front of a camera, they train a computer to associate certain poses as "left","right","up","down", and "neutral." Afterwards, they use those commands to navigate the maze. By allowing users to train their own commands, my hope was to make the experience more personal, to add a creative elements on the user's end, and to directly engage users in the process of machine learning. I used p5.js, ml5.js, and Arduino. 

Students and faculty tested out my project at Duke's Art, Art History & Visual Studies Department's End of Year Student Exhibition.