Test data creation
Just as we started our brainstorming session for our hack, we began by writing notes down to help keep out ideas in check. But of course like most of our other young and foolish peers we lost all of our papers as we went out to explore the inner workings of the cafeteria. Immediately we realized that if we were able to store the writing online by just using something unnoticeable and simple we would able to avoid stressing over not having lecture notes from professors who dont upload resources, losing notes you wrote on your hand as they wash away, and even phone numbers of girls who you finally got the courage to talk to! It would be the solution to oh so many problems and just another step further in getting it more digital.
What it does
As you write down notes on a sheet of paper, it uses an accelerometer to analyze your handwriting by returning the values returned from the accelerometer through a neural network that returns the most probable letter you are writing and maps it onto a text file you can download on your phone through an Android App. Most other solutions require buying expensive packs of paper that you'll probably lose anyway! But with this CHEAP, did I mention CHEAP add-on, you can get the best of digital and handwritten notes, on the CHEAP.
How we built it
We put an accelerometer on an arduino and wrote a CSV Writer to write the file to make a dataset of 100 training cases for each letter and 20 test cases for each letter as well as 350 features to each training case. Concurrently we built a neural network design for training the dataset that utilized a convolutional neural network and other layers that heleped optimize the performance of the algorithm.
Challenges we ran into
We wrote our entire program without being able to test until 7am , because of a lack of datapoints. This was a really big risk to us. The android application was also more difficult to write than expected, and required lots of reading to do. We also had some problems interfacing the android application with our computer, and had to fix it.
Accomplishments that we're proud of
It works and is possibly marketable. We also had lots of fun doing this.
What I learned
Damon Spencer and Kenneth Mitra and Austin Joseph: Android Development Karim Karim: Arduino Programming Austin Joseph: Machine Learning
What's next for Digiscripta
create a kickstarter Shrink Digiscripta down to fit on a pencap(Damon Spencer) Set up digital notetaking platform integration (Gdocs, evernote, onenote, etc.) (Austin, Karim, Kenneth) Find some investors to help us bring this to market
We would prefer the general track prizes over the IoT ones