Bugs / invalid code paths in Python API

Omniverse Isaac Sim
1

Hi everybody,

first off, thanks for the great tooling around Omniverse and Isaac Sim!

I’ve stumbled across a few code paths in the Python API which do not work or do not work as expected. I was able to work around these issues, so this post might be helpful for others having similar issues.

Data about my system:

Isaac Sim: 2021.1.1
OS: Ubuntu 18.04.5
System: Azure T4 (as suggested in the docs)

So here’s what I’ve found so far:

  • most ISAAC related modules should be imported after a object of OmniKitHelper has been instantiated
  • actually it is done correctly in the ISAAC Python samples, but the docs do not explicitly state that this paradigm should be respected
  • the problem for me was that when I imported e.g. from pxr import UsdGeom before a OmniKitHelper object was created, I got a few errors in the log output, but ISAAC Sim started just fine and seemed to be working
  • the issue got apparent when MDLs were involved in my scene, because they did not get loaded correctly anymore

=> I would suggest to emphasize this aspect (import after OmniKitHelper was created) more in the samples and docs

In $HOME/.local/share/ov/pkg/isaac_sim-2021.1.1/exts/omni.isaac.synthetic_utils/omni/isaac/synthetic_utils/scripts/synthetic_data.py

  • l.127, get_pose is missing a argument
  • get_pose is invoked in l.208 gt[sensor] = self.sensor_helpers[sensor](viewport) with viewport as argument
  • I would suggest to just add a unused argument to get_pose
  • however if that is done, the method crashes in l.134 prim_path = m[0]
  • I found that the prim_path is actually stored at index 1
    • print output was (1, '/World/fullmodel/rigid', 1, 'ilm', list( ... ), '')
  • similarly this applies to l.137 pose.append((str(prim_path), m[1], str(m[2]), np.array(prim_tf))), where class and label are stored at indices 2 and 3 respectively

Final snippet looks like this:

def get_pose(self, viewport):
     """Get pose of all objects with a semantic label.
     """
     stage = omni.usd.get_context().get_stage()
     mappings = self.generic_helper_lib.get_instance_mappings()
     pose = []
     for m in mappings:
         prim_path = m[1]
         prim = stage.GetPrimAtPath(prim_path)
         prim_tf = UsdGeom.Xformable(prim).ComputeLocalToWorldTransform(0.0)
         pose.append((str(prim_path), m[2], str(m[3]), np.array(prim_tf)))
    return pose

In $HOME/.local/share/ov/pkg/isaac_sim-2021.1.1/exts/omni.isaac.synthetic_utils/omni/isaac/synthetic_utils/scripts/writer.py

  • l.197 sensor_settings[name] = copy.deepcopy(sensor_settings_viewport) module “copy” does not exist (not imported)
  • this path only gets executed when the parameter sensor_settings with the value None (which is its default value)
  • if sensor_settings is not None, then there is no issue (that’s how it is done in the ISAAC Sim samples)

I hope this helpful to someone. In case of questions feel free to ask. I will also update this post if I come across other issues!

1 Like
2

Many thanks! We will feed this information into the development for our next release, coming up in a matter of weeks.

–Mike

3

Thanks for the feedback! We will take a look at the issues you mentioned with the synthetic_utils

The import order is briefly mentioned in a note here but we will work on making this more visible

https://docs.omniverse.nvidia.com/app_isaacsim/app_isaacsim/python_environment.html#omnikithelper

1 Like
4

Thanks, I totally overlooked that note! I actually never reached that “Omnikit Helper” section. Because the Option 1 - Default Python Environment section worked so well for me, I did not read past it