This chapter brought together a wide variety of tips and tricks concerning the cascade classifier object detection interface in OpenCV 3 based on the Viola and Jones framework for face detection. We went through each step of the object detection pipeline and raised attention to points where it can go wrong. This chapter supplied you with tools to optimize the result of your cascade classifier for any desired object model, while at the same time suggesting optimal parameters to choose from.

Finally, some scene specific examples were used to illustrate that a weaker trained object model can also perform well if you use the knowledge of the scene to remove false positive detections.