search

ros4hri / hri_fullbody

A ROS4HRI-conformant node to track and extract 2D and 3D skeletons of humans from RGB/RGB-D cameras
Apache License 2.0
14 stars 4 forks source link

Very small joint states numbers #2

Closed manuelgitgomes closed 1 year ago

manuelgitgomes commented 1 year ago

Hello all! First of all, thank you for developing this package, as it is very useful. I first hear of you from your presentation at ROSCON 2022.

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.

Kind regards, Manuel

LorenzoFerriniCodes commented 1 year ago

Hi @manuelgitgomes!

I'm happy to hear you're finding the package useful.

Regarding the small joint values: which ikpy version have you installed? Unfortunately, I noticed a similar behaviour in ikpy 3.3 (here the issue I opened some months ago), while ikpy 3.2.2 was working fine.

LorenzoFerriniCodes commented 1 year ago

Hi @manuelgitgomes, any update on this?