We begun this chapter by discussing the drawbacks of deep feedforward networks and how CNNs had evolved to overcome their drawbacks. Next, we dived deep into the architecture of CNNs, understanding the different layers of CNN—the input layer, convolution layer, maxpooling layer, and fully connected layer. We looked at the architectures of some famous image classification CNNs and then built our first CNN image classifier on the CIFAR-10 dataset. Then, we moved on to object detection with CNNs. We discussed various object detectors, such as RCNN, Faster-RCNN, YOLO, and SSD. Lastly, we used the TensorFlow detection model zoo to implement our first object detector using SSD.
In the next chapter, we will look at CNN architectures that require less computational power and are lightweight to run on a mobile device. They are called MobileNets!