Introduction

Convolutional neural networks are not only about images.

They teach a reusable idea: when data has local structure, a model can read small neighborhoods with shared weights. This creates detectors that can be applied across positions.

This chapter is bounded. It does not try to become a computer vision course. It teaches the vocabulary that later deep-learning discussions often assume: kernels, feature maps, padding, stride, channels, receptive fields, parameter sharing, and pooling.

By the end, you should be able to trace a small convolutional layer and explain why it uses fewer parameters than a fully connected layer over the same grid.