This chapter introduces Convolutional Neural NetworksConvolutionconvolutional [neural] network (CNNs), a classClass of neural networksNeural network specifically designed to process data with a grid-like structure, such as images. Unlike fully connected networks—which require input images to be flattened into vectorsVector, thereby discarding spatialSpatial relationships—CNNs preserve the two-dimensional structure of images and effectively capture localLocal pixel correlations. They achieve this by replacing standard matrixMatrix multiplications with convolution operations in one or more of their layersLayer. The chapter begins by presenting the general architectureArchitecture of CNNs and then details the components of a convolutional layerLayer, including the convolution, activation (detector), and poolingPooling stages. It also discusses strategies for initializing network weights. Several landmark CNN architecturesArchitecture, including LeNetLeNet, AlexNetAlexNet, VGGVGG, InceptionInception, GoogLeNetGoogLeNet, U-NetU-Net, ResNetResidualresidual network (ResNet), and DenseNetDenseNet, are introduced. Collectively, these topics provide a solid foundation for understanding the design and functionality of modern convolutional neural networksConvolutionconvolutional [neural] network.