---- How does it work? ---- This project uses a convolutional neural network, trained in Python. I imported all the weights and biases into Scratch lists and re-created the network. For anyone interested, this is the network architecture: Convolutional layer (3x3 filters, 6 output channels) Max pooling layer (2x2) Dense layer (30 neurons) Dense layer (30 neurons) Dense
