MNIST Neural Network

• 

  Example

Inspiration

For an organization at UT, I was required to do a capstone project. Since I was interested in machine learning, I decided to learn about neural networks, which led me to the classic MNIST project.

What it does

It is a neural network that is trained to recognize handwritten digits. The program trains the neural network using the MNIST training data, determines the network's accuracy using the MNIST test data (and prints it), and then outputs its prediction for a user-written digit.

How I built it

I wrote the program in python and trained the neural network with Tensorflow. The user-written image was read using Pillow.

Challenges I ran into

Tensorflow was difficult to install properly and the API was a challenge to learn. I had the most trouble regularizing the user's handwritten image to be fed into the trained network.

What I learned

I gained an understanding for what neural networks are and how they work. I also learned a fair amount about Tensorflow and how it works, as well as python and using a Linux machine.

Built With

• pillow
• python
• tensorflow