We tested out the Nuitrack AI SDK with INTEL RealSense Depth Camera D435, We tested out on multiple test subject but we are having some issues with the skeleton tracking and the depth,
The skeleton tracking sometime is difficult to track and when it does track the 3D model leg are somewhat not align correctly. The upper body seem to be fine but below body the leg area the tracking is not 100%, sometime its okay sometimes its not.
The depth as we can see here have a lot of noise and this make the tracking for skeleton hard due to lost of body features, and its hard as well when I’m trying to used the hand gesture because it does not detect the feature of my hand.
Some example of the Skeleton tracking and the Depth :
as you can see on test 1 - 3 we manage to track the skeleton but the 3d model leg its not align correctly, we followed the model rigged “Unity Chan” provided in the sample, and at first we thought we miss some step, but when we check back and run the sample provided in the Nuitrack sample, the same thing happen. The leg is not straighten, it bend inside a bit and sometime it just break the leg outward.
and for test number 4, the skeleton cannot detect the leg at all, it can only detect the knee area, but the knee area is at a wrong place.
Below is the SDK, pc spec and camera I used on development,
SDK:
Nuitrack AI SDK
Camera :
INTEL RealSense Depth Camera D435
Computer :
Dell G3 15 3500 Gaming Laptop
Windows 10 Home Single Language(64bit)
NVIDIA® GeForce® GTX 1650 Ti 4GB GDDR6
12GB DDR4-2933MHz, 2x4G
512GB M.2 PCIe NVMe Solid State Drive
So from all of the example above, is there a way for me to improve the skeleton tracking and fix the 3d model leg so it does not bend inward and break. Is there any documentation regarding on how to rig the 3d model so it can be used properly with Nuitrack skeleton tracking.