[SOLVED] Pointcloud generation in C++

I am currenlty working on a Project where i need to export the Skeleton Joint Data and the corresponding Pointcloud form the Depthframe data in C++. I tried to follow the Pointcloud Tutorial in Unity but sadly wont be able to implement the methods required in C++. I am able to export the joint data without any problems but when it comes to exporting the Depthframe data as a pointcloud i am unable to find any Functions that could be of any help.
Has this been asked before or is there an implementation for the Function that i couldnt find?
Any help would be much appreciated.

Hi @tbasit,

There is no ready to use function which would return you a point cloud. But you can implement your own point cloud generation using depth information which is provided by Nuitrack.

  • First you need to get a depth and RGB frames (for example you can look how to get depth frame in nuitrack_gl_sample
  • Second you need to convert each point of the depth frame to point in real world coordinates (XYZ). That could be done using convertProjToRealCoords function.
  • After you do that, you can use those coordinates to render corresponding point of RGB frame using rendering engine of your choice.

Thank you for the fast reply. I tried your suggested method but i have a problem with the convertProjToRealcoord as it accepts three arguments x,y and a Depth Map.
I have tried multiple methods which should be outputting a depthframe but none of them seem to work as the third input parameter to the above stated Function. I also tried to use the depthSensor->getDepthFrame() method and sadly that is also not the right data type.
The nuitrack_gl_sample does it the same way as far as i have seen other than the step with texture mapping.
Am i missing something?
Can you please refer me to the Function which delivers the right output.


Hi @tbasit,

convertProjToRealcoord needs not DepthFrame object, but actual data of depthframe as the third input parameter. So you would have to call DepthFrame::getData() to get an array pointer and then feed certain values of that array to convertProjToRealcoord as the third input parameter.

const uint16_t* depthPtr = depthFrame->getData();
Vector3 realCoords = _depthSensor->convertProjToRealCoords(x, y, depthPtr[y * depthFrame->getCols() + x]);

Hi @tbasit

Has your issue been solved? Do you have any other questions that I can help you with?

Hello @a.bragin
The Problem i ran into after your suggestion was as follows:
As i need to call the convProjToRealCoords function inside the onNewDepthframe callback i am having problems in passing the Depthsensor object to the callback function. Is there any other way to call this Function without first instantiating a Depthsensor object and passing it.
My main goal is just to get a Pointcloud and the corresponding Skeletal joint data for use in my project.
For that i need the Points to be in the same Coordinate system the previously suggested implementation should theoreticaly Work but as stated above i was unable to get it working as intended.

Hi @tbasit,

You can pass Depthsensor object to the callback function using std::bind
E.g. depthSensor->connectOnNewFrame(std::bind(depthCallback, std::placeholders::_1, depthSensor))

Hi @tbasit

Has your issue been solved? Do you have any other questions I can help you with?