This project is real-time visualization of a network recognizing digits from user's input.
I trained a network using MNIST dataset and parsed the weight data in python. With this data, I implemented my own custom functions that are needed to run the network in Processing including matrix multiplication function, activation functions. At first trial, because MNIST dataset is preprocessed for numbers to be in the center of the images, there was a precision problem when the user's input is placed little bit far away from the center. I used data augmentation technic in the training process to resolve this problem.
To run this code, you need Processing IDE and a library named peasycam.
Subscribe to Python Awesome
Get the latest posts delivered right to your inbox