sorry if this has come up before: is it possible to connect to multiple cameras, each with its own trackers? For our project, we need a very wide field of view, so we are installing 2 RealSense D435 cameras. It would be great if we could reliably connect to both of them (fully aware of which one is which) and run user tracking on each one independently, on its own thread if necessary.
I guess a secondary question would be: how to specify which camera to use if there are more than one connected to the PC? Is this a matter of supplying a different config file to each instance of Nuitracker?
Multiple cameras is not possible with the current intel cameras - its not a supported firmware option last time I checked. They are pretty agressive when it comes to USB bandwidth also FWIW. and would likely need entirely separate usb 3.1 controllers and serious processing power to attempt it.
I would suggest you may need to look at some form of dual computer setup - with tracking data being passed around over IP or some such
Its basically not possible to tell nuitrack what sensor to use in its current form. When you run INIT it does a search and uses the first depth device it discovers. There is an option for RGB ids with the ORBBEC Pro - but thats not controlling the depth sensor selection.
thanks for your reply. This may mean we’ll have to connect the cameras to their own little NUC or something, which is a scenario I had in mind before being told the producers would rather not. So far, I have not been able to run 2 cameras at the same time, but this may have been a cable problem. New high speed cables have been ordered. Running one camera on a USB 3.0 (Gen 1) hub did not pose a problem, especially not at reduced resolution and with the color camera disabled.
Regarding the Nuitrack API: it might be wise to give users a little bit more control over how to connect to a camera, perhaps by offering a way to list the connected supported devices and accepting an id parameter on creation of the depth sensor.
Update: the two cables arrived. I am now able to run both cameras from the same laptop using the RealSense Viewer application over USB 3.0. System is an Intel i7-4720HQ CPU @ 2.6Ghz with 16GB RAM running Windows 10 64-bit. Estimated CPU usage when running both cameras in depth-only mode @ 848x480 is 25%.
Anyway, I’m marking this topic as “solved”. Thanks for your input.
The challenge is still that you cant currently reference the two sensors inside NUITRACK.
Yeah, that’s a dealbreaker. Shame, really. To me it sounds like a simple change to the API is all that’s needed to make it possible.
The issue for us is the current Init() system. There is no way of controlling which device the SDK initialises.
We often have multiple devices and types connected to test systems … and really at a minimum need a way to be able to say - USE this device.
At the moment its not even possible to determine which device was even selected. That is also high on our wish list. A way to ask the SDK what device make and model is being used currently.
In that case I think Intel should take-over NuiTrack. Problems will be solved automatically.