Object (or target) tracking plays a key role in surveillance systems, autonomous vehicles and intelligent transportation systems. Classical model-based multisensor-multitarget tracking techniques such as the Multiple Hypothesis Tracker (MHT) and the Joint Probabilistic Data Association (JPDA) filter rely on target motion and sensor measurement models to form and maintain tracks. The resulting tracks are often used learn more about target types, behavior, intent, etc., for situational awareness. In contrast, emerging data-driven machine learning (ML) or artificial intelligence (AI) based techniques rely on extensive a priori datasets to learn target models, which in turn are used to track targets and their behaviors. Future operational systems are expected to be hybrid in that they will rely on both model-based and data-driven techniques. This presentation explores the interplay between learning and tracking, with one leading to the other. We look at some representative problems where tracking is used for learning and vice versa. We also discuss performance evaluation and quantification, ground-truthing and seamless integration of both types of techniques. Real data from airborne surveillance and ground-based autonomous driving systems are used for demonstration.