Master Moodway

Ever take a gaming session too far? Even the calmest among us know that sometimes, anger can get the better of you while trying to have fun. Enter Master Moodway! The wise turtle is here to calm you down and stay focused on what really matters.

Master Moodway grew out of our team’s common experience as gamers. Fitting in with the arcade game theme, we wanted to make something that could check you while you play, since it’s easy to get lost in emotion. We built the project using Python and various AI models to analyze facial expression and sound, while we used Javascript to make the tiny display with Master Moodway. The visual model comes from DeepFace, and the audio model comes from wav2vec2.

This did not come without its challenges. One big challenge was trying to connect the backend to the frontend using socket-io, which is built on top of Web Sockets. We had to work with managing dependency compatibilities and using similar sounding methods from different modules (sio.sleep, eventlet.sleep, time.sleep, etc.). Another challenge we faced was determining the weights and thresholds to reasonably detect enough “anger” from the models. Through trial and error, we settled on a set-up that satisfied us. A quirk we discovered was that the audio model we were using associates a muted mic with anger. While at first we worried how to fix it, we realized we could use it as a testing scheme.

Through this project, we learned a lot about how to use Web Sockets and integrating various AI models into one project. Also, we learned how to implement a kind of streaming API rather than the traditional REST model. Generally, we got exposure to more libraries. Through these things, we created something we can be proud of.