Summary

In this chapter, we introduced CNNs. We talked about their main building blocks – convolutional and pooling layers – and we discussed their architecture and features. We paid special attention to the different types of convolutions. We also demonstrated how to use PyTorch and Keras to implement the CIFAR-10 classification CNN. Finally, we discussed some of the most popular CNN models in use today.

In the next chapter, we’ll build upon our new-found computer vision knowledge with some exciting additions. We’ll discuss how to train networks faster by transferring knowledge from one problem to another. We’ll also go beyond simple classification with object detection, or how to find the object’s location on the image. We’ll even learn how to segment each pixel of an image.