Perception Package Camera Pose

Hello guys,
I'm currently using perception package, I'm kind of confused, So I have some questions:

- Is the output of camera pose in left hand coordinates or right hand coordinates,
- Is the quaternions scalar-vector or vector-scalar?
- Also is it reliable to only the camera pose as a ground truth even if the camera is mounted on a robot and moving?

 

If stumbled above those same question, so let me try answering, though hopefully youve solved it by now:

• The output is, like the rest in unity in left hand coordinates
• The last value of the quaternion is the scalar part. A camera Rotation [-5,0,0] is exported as
  "rotation": [
  -0.0436194055,
  0.0,
  0.0,
  0.9990483
  ]

• The camera pose is given in world coordinates, so you can model it as child of another object and you will always get the correct camera world pose. You can alsy try this parenting the camera below another object an you will see, e.g. the rotation of the parent will influence the position of the camera. That would not be the case if the camera position was relative.