An example provided by the lovely Berk
An example of the processing of the images
There is a severely lacking market in the world. It is practically impossible for average individuals to Interpret handwritten numbers. We wanted to help those in need by providing a method for digitizing these values.
What it does
Using a complex neural scheme inspired by nature, we theoretically have the capabilities to train our network using a data set of over sixty-thousand handwriting samples. However, we have found that even without input, our network is so highly sophisticated that it can train itself to map to the correct outputs. Once training is compete, newly written samples can go through our half-hour, semi-manual image pre-processing pipeline. This pipeline is unlike anything the world has ever wanted to see. After pre-processing is complete, we can pass this image into the network and receive a digital numeric output. Upon passing through the image, the neural network will inform you that it was, in fact, a three.
How I built it
Working with our engineering staff and Aaron Aaeng, our Chief Visionary, we constructed the network in Python, a cutting edge programming and software development language. There are a series of input neurons that take the gradient value of the input image. This is passed through a series of dynamically optimized activation functions through a hidden layer. These hidden nodes map data to determine the value of the input image.
Challenges I ran into
There was a major problem with the implementation of the unpickling pipeline. Solving this issue almost brought the end of our Chief Visionary. This problem was so grand that though we registered as a team of four members we are completing today a team of just three. Unfortunately, this team member was the only one that had an actual idea. However, we collectively refused to attempt this idea. Also, there was this one partial derivative that was too long to be listed in the graduate level textbook in which we found it, so that was a little difficult.
Accomplishments that I'm proud of
I am proud of Ethan for constantly saying that he solved the problem, then clarifying that overall functionality has decreased. It takes true personal confidence to lie to your group with such frequency. I am proud of Aaron for never giving up even as his CPU literally melted off part of his laptop as he became an expert on Pickle. I am proud of myself for not physically harming Aaron throughout the course of the event.
What I learned
I learned that though we were close to being a complete development team, our cooperation somehow makes us less than the sum of our parts. We realized that had we used Slack for instantaneous communication, it would have optimized our ability explore user stories. We realized that had we worked using pair programming, Ethan would not have been able to, for the first time in human history, achieve a segmentation fault in Python. For those of you that are wondering, it was a method convert between one and two-dimensional lists.
What's next for Num
Next, Ethan is going to dedicate the next week of his life to finishing this project to prove that he had it in him the whole time.