Description of a Tracking Metric Inspired by KL-divergence
share
A unified metric is given for the evaluation of tracking systems. The metric is inspired by KL-divergence or relative entropy, which is commonly used to evaluate clustering techniques. Since tracking problems are fundamentally different from clustering, the components of KL-divergence are recast to handle various types of tracking errors (i.e., false alarms, missed detections, merges, splits). Preliminary scoring results are given on a standard tracking dataset (Oxford Town Centre Dataset). In the final section, prospective advantages of the metric are listed, along with ideas for improving the metric. We end with a couple of open questions concerning tracking metrics.
READ FULL TEXT
