The architecture of LeNet-5 includes the C1 layer (6 5×5 convolution kernels, output 6@28×28), the S2 layer (2×2 average pooling, output 6@14×14), the C3 layer (16 5×5 convolution kernels, output 16@10×10), the S4 layer (2×2 average pooling, output 16@5×5), the C5 layer (120 5×5 convolution kernels), the F6 fully connected layer (84 neurons), and the output layer (10 nodes).