where true positive (TP) is the number of correct detections of the ground-truth boundingbox, that is, the number of IoU that exceeds the threshold and is correctly classified; false positive (FP) is the number of incorrect detections of a nonexistent object or misplaceddetections of an existing object, that is, the number of IoU not exceeding the threshold orthe number of misclassification errors; and false negative (FN) is the number of misseddetections, that is, the number of boxes that are not predicted [47]: