Question about Considerations in Skeleton Detection

When detecting a human skeleton through Nuitrack, can the accuracy of the result vary depending on the target’s race, age, height, body type, disability, etc.? In other words, are the above variables taken into account in the results Nuitrack detects?

This article was translated by Google Translate.

Hi @wpjeon,
in short - it depends:
"Classical" skeletal tracking engine uses mostly depth data for tracking so it doesn’t depend on factors related only to RGB image (such as race and age for example).
Nuitrack AI tracking engine also relies on RGB data so it indirectly depends on most of the factors.

Dependency is somewhat complex and particular use-case should be tested specifically.
In general guideline could be formulated as following:

best tracking results are achievable for :

  • adult person
  • of average height 160-190cm (5’3 to 6’3)
  • standing frontal to the camera.

each discrepancy (in terms of base factors) could impact tracking performance

Please also note that it is possible to improve tracking for any particular use-case through this tool, so please don’t hesitate to contact us.

1 Like

Hi @wpjeon, how are you ?
It would be great if you could share some feedback (does Nuitrack work for your needs?).

1 Like

Sorry for the late reply. Your answer was very helpful. I think We can refer to your answer when developing. Thank you again.

Thanks. Please feel free to ask in case of any questions / issues.