We are heading towards the end of this chapter, but before we end, I would like to address two small but still important topics. Let's start by discussing the evaluation of the performance of cascade classifier object detection models by using not only a visual check but by actually looking at how good our model performs over a larger dataset.
We will do this by using the concept of precision recall curves. They differ a bit from the more common ROC curves from the statistics field, which have the downside that they depend on true negative values, and with sliding windows applications, this value becomes so high that the true positive, false positive, and false negative values will disappear in relation to the true negatives. Precision-recall curves avoid this measurement and thus are better for creating an evaluation of our cascade classifier model.
Precision = TP / (TP + FP) and Recall = TP / (TP + FN) with...