Summary 

You are still here. I'm impressed, and congratulations! It was a long but fruitful chapter. We saw another example of how we can apply CNNs, and in doing so, we further developed our understanding of how they work, how to tune them, and ways in which we can modify them. We saw how we could use the learned features not just for classification but ranking, a technique used in many domains such as fashion discovery and recommendation engines. We also spent a significant amount of time building a drawing application, which we will continue to use in the next chapter. There, we will again explore how to perform sketch classification using a RNN trained on Google's QuickDraw dataset. Lots of fun ahead, so let's get started.