@dusx
Hi Ryan,
Thanks for the update. I get that about Orbbec SDK development but they appear to be leading this area of tech.
Something that is very critical for our work in media and performance is the ability to access skeleton tracking in low light, highly dynamic and no-light situations. That is why Kinect style tracking is still relevant despite its age. As you know, the structured light camera sensors (Kinect/astra pro) emit their own IR light source and are mostly agnostic to other light sources and projections. This is a critical quality for media and performance work, wouldn’t you agree? The additional parameters available with OpenNi - depth calibration, single or multiple body isolation - provide methodology in creating logic for real-time body tracking installations (Where I am at with this current project).
ML body tracking works with RGB video streams really well. The option to implement MediaPipe style ML body tracking is already an option by pairing Isadora with 3rd party apps that output body tracking data as OSC or json. But these solutions, while great, do not function well in low light and, moreover, will annoyingly track unwanted bodies appearing in projections or screen media in the real-time capture environment.
There might be some short term potential to pair depth camera sensor sources with a Python based ML body tracking solution in Isadora, but this would require access to depth cameras and a Python module in ARM version, both requiring development time and therefore prioritisation by yourself and @mark.
This brings us back to body tracking with the ancient OpenNi solution - particularly when it is the public audience/participant that we are tracking to create an interactive work in a darkened gallery setting.
In terms of a dedicated Orbbec module - using depth data with blob tracking is useful and worthwhile. I guess if we could do something with point cloud data in Isadora would be a game changer.
For now though, correct me if I am wrong, the OpenNi Tracking provides the most efficiency for real world application in media and performance in terms of working with Isadora. So we are thankful that it is still supported and functional in the current environments.
Best wishes
Russell