Using Nuitrack for recorded .bag file

Hi,

I want to use Nuitrack models on a collection of recorded .bag files from Intel RealSense cameras.

What is the best platform and SDK for this objective?

Best,
Koorosh

I tried Python SDK with setting “Realsense2Module.FileRecord” in the config, but it requires activating the licence while the camera with the licence enabled is connected to the computer.

Hi @KooroshRH,

it’s correct, currently Nuitrack Trial license is needed to be able to perform tracking on pre-recorded sequences (*.bag files is one example).
To activate Trial license you need one of supported sensors. Do you have any ?

Hi @TAG,

Thanks for your response.

I have enabled a perpetual licence on one Intel RealSense camera. It works with the desktop app and the Python code while using live capture from the camera, but when I keep it connected and want to use .bag files as source it returns this error after showing some frames of the .bag file:

nuitrack.update()
RuntimeError: Nuitrack update failed: License not acquired

Hi @KooroshRH,

Sorry, but just have to re-cite previous message “currently Nuitrack Trial license is needed”
Perpetual license doesn’t work in this mode, only Trial.

Let me explain a little bit: currently processing of pre-recorded sequences isn’t an official feature, it’s just a semi-documented testing tool which is provided as-is (it supports only small subset of depth sensors). That’s why it is available only with Trial license and with a 3-minutes limitation.

We’re thinking about changing this (in fact for a very long time), but have a limited demand for this.
Does your project rely on this as a core function without current technical limits or it’s also just for testing purposes ?

I understand, but actually I counted on this feature and recorded many .bag files to do the skeleton tracking using the Nuitrack in future. Also, I am considering doing the multi-device from different viewpoint skeleton extraction as well.

Hi again @TAG,

Is it possible to have a trial licence to at least try this feature to process pre-recorded bag files? I checked the portal and I couldn’t find a way to get a trial licence.

Trial license is available at the licensing dashboard

“Main page” (Nuitrack should be selected in side bar) → “Choose a license”

Hi @TAG,

Thanks for your help. I was able to run the model on pre-recorded files, but I’m trying to fine-tune my experience. I need to use the user tracker part of the model, but it is not clear how I can use it using the Python wrapper. Is there any documentation for the Python wrapper so I can read it?

Also, isn’t is possible to enable the pre-recorded feature for perpetual license? Since it seems working as I expect using the trial license but with the duration limitation. Some of my recordings are lengthier than 3 minutes.

Hi @KooroshRH,

  • currently our python package is still in beta and is lacking user masks / segments feature. Do you consider using other APIs (C++) or you’re really bound to python as primary tool in your project ?
    (We’re asking as our python package is falling behind feature-wise just because of limited understanding of use-cases)

  • as was said Nuitrack Perpetual currently doesn’t work in this mode
    Just as a consideration for future versions - would it work for you if Nuitrack AI subscription will support recordings processing?

Hi @TAG,

I can simply switch to C++, but my preference is Python. If there is no firm documentation on Python package I will switch to C++.

It will absolutely work for me if Nuitrack AI can support recordings, since using it while collecting data can cause overheads for the data collection process, so it’s better for us to record everything in .bag files and process them in future.

It will absolutely work for me if Nuitrack AI can support recordings

Important part in the question was about subscription - “would it work for you if Nuitrack AI (annual) subscription will support recordings processing”, as we currently can’t see an option of supporting this mode for Nuitrack AI Perpetual license.

Hi @TAG,

Yeah sure that works as well, but does that mean we need to buy another subscription which is annually and not unlimited?