Digit Recogniser with Convolutional Neural Networks (CNNs)

CNN Architecture

CNN architecture is adapted from the VGG architecture.

1. Block 1 and 2

• 2 convolutional layers
• 2x2 max-pooling layer
• 25% neuron dropout

2. Block 3

• Dense layer of size 128
• 50% neuron dropout
• Dense layer of size 10 for each number