Now, when I tried to use the package, I cannot get a correct estimation of a 2D/3D skeleton pose using an RGB-D camera.
One thing I noticed is that most values for the joint states, as they are really small (near zero):
position: [0.0, 0.0, 0.0, 0.0, 0.0, 8.909208311087425e-11, 0.0, 1.631412601457227e-10, 0.0, 1.128599690590957e-10, 0.0,
1.1872098531808344e-10, 0.0, 6.628555182943805e-11, -1.8735825248397015e-11, -1.6691366356208623e-10,
-0.09999942649329777, -0.39999999999998326, 0.6999999999999998, -3.535952186191125e-18]
These values, however, move consistently when the real world joint moves. That is, in a repeating joint movement (rotation of left arm, as an example), the joint values associated follow the movement. These joint values continue to be very small (in the same order of magnitude as the ones presented before).
To tackle this, I have tried to multiply these values by a large constant, with no success. Do you have any suggestion on what might be wrong?
I have a bagfile being uploaded to help you test in my conditions, if you want to.
Thank you for your attention.