Heuristics is essentially codifying rules you have in your head to solve a particular task, typically implemented using a function that returns a score. This is used to rank a set of alternatives. In the previous section, Designing in constraints, we saw how we could use the center of mass and the bounding box to determine how well distributed the pixels are for a segmented image. This in turn could be used to rank each frame, by favoring those with a center of mass near the center of the bounding box. We also implemented a type of heuristic in the application when determining which frames to keep and which ones to ignore by measuring the overlap. 

Heuristics can be a powerful ally, but be careful to ensure that the heuristics you derive can generalize well to your problem, just as you would expect a good model to. Also, be mindful of the additional computational cost incurred from using them.