CNN for QMNIST

In this project we train a CNN for handwritten digit recognition. Read Convolutional Neural Networks and CNNs with Keras before you start.

You may reuse code from Load QMNIST and Image Processing with NumPy.

Task: Solve the QMNIST digit recognition task with a CNN. Try to get 99 percent correct classifications on the test set (without using the test set for training, of course). Remember that convolutions give more weight to image center. Thus, do not crop QMNIST images. Use 28x28 images with centered bounding boxes. Save your model to a file.

Solution:

# your solution