Issues whilst using 3D multi-camera inputs

[timkoers]

Hi!
I'm having issues using the multi-camera setup through Python.

I've taken a deep dive into the source to try and see what goes wrong, but I could not determine it yet.

I'm emplaceAndPop'ing a VectorDatum into the OpenPose wrapper, containing two datums, each with their respective cvInputData and cameraMatrix.

However, when running openpose, I'll get the error that when using --3d mode, I should use two cameras (https://github.com/CMU-Perceptual-Computing-Lab/openpose/blob/master/src/openpose/3d/poseTriangulation.cpp#L221).

I do not understand how the flow should be as

wrapper.hpp receives the VectorDatum, it'll split it up into individual datums that are getting fed into the threadmanager, which in turn call the poseTriangulation thread for pose triangulation. This means that poseTriangulation will never receive two datums, as they are always split up (https://github.com/CMU-Perceptual-Computing-Lab/openpose/blob/master/include/openpose/wrapper/wrapper.hpp#L535).

Please explain how I should provide the data to allow for multi-view pose estimation