Initially we plan to build random music generator that combines LSTM into it to build a random music using magenta.js and it turned out that magenta is so hard to use at this phase, and we finally turned to vlc media player.
What it does
It basically detects several of your emotions in real time, it can be done in using picture as well. And then, it can detect the gender! THE music player! Also, based on your facial expression, it samples the information for a certain period and plays the corresponding music for your mood.
How we built it
We play and toy around with CNN, getting the dataset by accident (just simply get it. Only god knows), and transfer learned the model We also prepared a tiny music database(of course expandable, but this one is just for demo) for you to play around with it.
Challenges we ran into
Dude, it is a lot. Working on this where every participants are doing AI related projects and turns out we are doing the same thing, made me feel a bit bitter :( Also the dependencies caused tons of problems.
Accomplishments that we're proud of
at least, we have done something that's working!
What we learned
Learn? learning how to stay awake 24 hours building project is something that I learnt.
What's next for emotion facial expression based music player?
Larger music database, more expression categories, more robust detections, faster running, variable sampling rate, a nice UI design, a full releasable APP!