digit This is a kaggle competition. I found an easy way to get 1.00 just by using the official MNIST data. You can use the .csv which I have merged to train.