My personal version of my team's project (called Num) for the 2017 T9Hacks Hackathon at CU Boulder. It is an (almost) from-scratch Neural Network for recognizing handwriting, trained on the MNSIT dataset. While the project won best marketing, there were a lot of unresolved issues which prevented the network from successfully recognizing handwriting. In this version, I rewrote the network class over a day or two and in the process fixed a lot of these issues. While there are still problems that I have not solved, the network now achieves close to 90% recognition.

Original Readme Below:

Who knows

A nueral net trained on the MNSIT dataset to read handwritten digits. Also has image processing to read new images, but there's no integration between the two modules. Won best marketing at the CU Boulder T9 Hacks hackaton in 2017. Project mmmmiiggghhttt be a bit of a trainwreck.