The idea behind FocusOn is based on a DJ created using AI that would be able to control the mood of a party based on cameras and sensors.
The original idea was mutated then into an assistant to the car, at the moment that we saw that this kind of application could work better as an assistant that as a DJ
What it does
The application control the user behavior based on a camera and the car sensors. In addition, it interacts with the user by asking for route changes or to stop or slow down the vehicle in situations of risk.
How we built it
The application is divided into a frontend, built as an Android app, and a backend built with Django.
In the case of the Android application the app is built with Java, using Google Maps API to provide a user interface.
In the case of the Django server we have Scikit-Learn to run the different models, OpenCV to process images and calls to API like Spotify or Azure
Challenges we ran into
Bosch IoT Lab
Accomplishments that we're proud of
- Get images in second plane using Android
- Train a model able to tag songs based on subjective things like anger
- Play with different API in a single project, like Spotify, Bosch IoT Lab and Google Maps
- Use a relevance feedback system as Rocchio
- Make the full development of an Android app, considering that we are not android developers
What we learned
- Android Development
- Machine Learning
- API usage
What's next for FocusOn
- Add more interactions with the user
- Expand capabilities