Is Object Detection more accurate than BodyTrack in positioning?

What is the difference between the two in terms of object positioning?

I noticed that the two have different instance_module_id when they are executed at the same time, which means that the data of the two are independent?

If object detection is better, I hope to use it for positioning, because when I use bodytrack, there are often problems such as jitter and inaccurate edge positions.

However, I need to use skeleton data, so how can I merge the two independent data together.