Digit classification using PyTorch.
Run vanilla_nn.py to get a trained model.
Load the model in draw.py, draw your own digit and classify it.
mnist_cnn is a basic convnet. To load this model into draw.py you'll have to change the class structure and change
line 68 to
img = np.reshape(img, (1, 1, 28, 28))