This code trains a convolutional neural network on the MNIST dataset. Uses the Keras and NumPy libraries.
INPUT -> [CONV -> RELU -> BATCHNORM -> POOL]*2 -> FC -> RELU -> BATCHNORM -> DROPOUT -> FC
This model has achieved 99.28% test accuracy on the MNIST dataset with batch size of 128 and 12 epochs.
sudo pip3 install -r requirements.txt
python3 mnist_conv.py