Binary Image Classification

Binary classification is the simplest approach for classification models as it classifies images into just two categories. In this chapter, we started with the convolutional operation and discussed how you use it as an image transformer. Then, you learned what a pooling layer does and the differences between max and average pooling. Next, we also looked at how a flattening layer converts a pooled feature map into a single column. Then, you learned how and why to use image augmentation, and how to use batch normalization. These are the key components that differentiate CNNs from other ANNs.

After convolutional base layers, pooling, and normalization layers, CNNs are often structured like many ANNs you've built thus far, with a series of one or more dense layers. Much like other binary classifiers, binary image classifiers terminate with a dense layer with one unit and a sigmoid activation function. To provide more utility, image classifiers can be...