We chose to help simprints by classifying fingerprints automatically to reduce search times.
What it does
Our idea was to use machine learning to infer from the training data how to fit images of fingerprints into 5 separate categories.
How we built it
We used sci-kit learn in python 2.7
Challenges we ran into
Individual pixels are not the best input to a machine learning algorithm, ideally there would be lots of preprocessing and dimensionality reduction, but we didn't have the time or expertise to do this.
Accomplishments that we're proud of
Learning very quickly and coming up with a program that uses a library we had never used before, in a language some of us had never used before
What we learned
That things are still very complicated even with a library - it won't do everything for you
What's next for Fingerprint Finders
Improve the accuracy of our algorithm somehow (we have done a bit of research into general machine learning techniques but not implemented them yet). We would like to reduce the run-time of our learning algorithm so that we can run it on the whole data set instead of just a small subset.