After seeing the MBed chip, we knew we wanted to do something with it. Ideas ranged from electronic limbo to smart smoke alarms. Eventually we settled on a sleep assistant as we thought it was an issue not normally touched on by IoT devices
What it does
The MBed chip posts data to a flask server once every two minutes in normal use, this is then processed everyday by a long short-term memory recurrent neural network which takes data from the FitBit API about the sleep quality and predicts the ideal environment for the user. All of this data is then shown on a React.js website so the user can easily see how they can adjust their sleep environment.
How we built it
The MBed chip runs on C++ and calls a flask server hosting a RESTful API. This is then used as an access point for the neural network and the front-end which both request data.
Challenges we ran into
It was our first experience with designing our own RESTful API which was a challenge to begin with as we had to learn lots of new packages in a short amount of time. As well as this, we hadn't created a neural network using our own data which raised it's own problems. Parsing data also became an issue towards the end as we hadn't though of all of the edge cases for dates and time.
Accomplishments that we're proud of
Creating a functioning RESTful API was a great accomplishment as it required lots of things to work together, also creating a responsive website near the start of the hack was amazing! Overall however, I think that the best accomplishment was not feeling rushed, through planning and distribution we managed to work together and only had a small rush at the end for submission.
What we learned
Python is amazing...most of the time.
What's next for DeepSlumber
Fully implement the learning algorithm and get some actual data rather than spoof data (you can't get lots of overnight data in one night!)