LeNet (CNN)

Yann LeCun, 1989

Developed by Yann LeCun in 1989 for handwritten digit recognition, LeNet demonstrated that convolutional neural networks could automatically learn spatial feature hierarchies. The visualization shows data flowing through the architecture: an input grid processed by sliding convolution kernels (gold), pooling layers that reduce spatial dimensions, and fully connected layers producing the final classification. Each layer block is sized proportionally to its dimensions.