In this chapter, we learned about SVMs in all their forms and flavors. We now know how to draw decision boundaries in 2D and hyperplanes in high-dimensional spaces. We learned about different SVM kernels and look at how to implement them in OpenCV.

In addition, we also applied our newly gained knowledge to the practical example of pedestrian detection. For this, we had to learn about the HOG feature descriptor, and how to collect suitable data for the task. We used bootstrapping to improve the performance of our classifier and combined the classifier with OpenCV's multi-scale detection mechanism.

Not only was that a lot to digest in a single chapter, but you have also made it through half of the book. Congrats!

In the next chapter, we will shift gears a bit and revisit a topic from earlier chapters: spam filters. However, this time, we want to build a much smarter one than in the first chapter, using the power of Bayesian decision theory.