What is this?

InsideOut is a machine learning mood ring. Just like the 50 cent rings that you can find in gift shops, InsideOut can tell you your mood---even if you don't know it! Simply plug in an arduino with leds wired to pins 2 through 8, train the machine learning model, and run the program! Using your webcam, InsideOut will scan your face and reveal your deepest secrets (well, maybe just if you're feeling one of the 7 emotions it supports).

Using Haar cascades to identify faces, then cropping those faces and feeding them into a custom convolutional neural network, InsideOut identifies the emotions of every face in view. The CNN was trained on a dataset of about 30,000 faces linked in the README.md file (which also contains instructions for installation and running). Each sample came with a 48x48 greyscale image of an emotive face and a number that identified one of 7 emotions.

What inspired InsideOut?

Everyone in our group was fascinated by computer vision, especially when it's used on humans. Emotions are extraordinarily complex, yet we seem to understand them with ease. This makes them a great candidate for machine learning. Once we started thinking about recognizing emotions, we realized that we had the chance to build a real mood ring, with actual data behind it. Unlike toy mood rings, this one actually works!

Built With

Share this project:

Updates