Table 5.

Effect of $\delta$

Model variant	Matching strategy	Perfect det.
		mAP	MOTA
$\delta =1$	IoU	58.64	39.03
$\delta =2$	IoU	54.71	38.42
$\delta =3$	IoU	56.66	39.31
$\delta =4$	IoU	56.35	38.09

Each model is trained with a separate $\delta$ value

The bolded numbers represent the best performing scores, in comparison between method