CNNs are the de facto image classification models due to their ability to learn the distinctive features of each class by themselves, without deriving any relationship between input and output. In this chapter, we learned about the components of CNNs that are responsible for learning the image feature and then classifying it into predefined classes. We learned how convolution layers stack on top of each other to learn from simple shapes (such as edges) to create a complex shape (such as an eye) and how the dimensionality of the feature map changes due to the convolution and pooling layers. We also understood the functions of the nonlinear activation function, Softmax, and the fully connected layers. This chapter highlighted how to optimize different parameters to reduce overfitting issues.
We also constructed a neural network for classification purposes and used the model...