Shapenet converter: duplicate file collision_geometry.obj in exported archives

[gabriel-v]

After rebuilding the shapelib container and starting the parfor script, I observed something off with the archives:

There are 2 files with the same name.

When extracting, one of the files overwrites the other one.

Here are the offending lines:

kubric/shapenet2kubric/convert.py

Lines 296 to 299 in a3d5eaa

	tar.add(object_folder / 'kubric' / 'collision_geometry.obj',
	arcname='collision_geometry.obj')
	tar.add(object_folder / 'kubric' / 'model_watertight.obj',
	arcname='collision_geometry.obj')

I suppose the second file should have arcname='model_watertight.obj', correct? I can't find documentation for the specific format this is converting to; I assume it's using the same format as the MOVi objects?

[MrXandbadas]

From what I've read in the code and here online your assumption on the name convention is correct.
When we add an object we need to pass through the noted two .obj files (both a collision_geometry file and a model_watertight file)
Try changing line 299 to arcname='model_watertight.obj' and let me know if it succeeds, i'll happily make a quick PR for someone to review if it does work

---

Edit (some Context I found:
#185
Qwlouse said:

To make this work for simulation you would need to create a collision mesh in obj format and a corresponding urdf file that you can pass as simulation_filename. If your mesh is watertight, you can take a look at the GSO preprocessing script about how to do that. If not, then things get complicated, and you will have to do more than that. The ShapeNet preprocessing scripts can serve as a starting point there.

Looking into the ShapeNet processing linked we see:

kubric/shapenet2kubric/convert.py

Lines 88 to 90 in 9dd4ed1

	def stage1(object_folder: Path, logger=_DEFAULT_LOGGER):
	source_path = object_folder / 'models' / 'model_normalized.obj'
	target_path = object_folder / 'kubric' / 'model_watertight.obj'

Edit edit -- Woah brain broke, what is this

[gabriel-v]

I made that change locally for line 299 and the archives look good. I have yet to verify them in the renderer, through; but the conflicting filename bug is solved by that change.

[MrXandbadas]

Instead of spam the PR section I've added it into the current fork I have listed in the Pull Requests xD

[Qwlouse]

That is an excellent catch!
I'll have to regenerate the ShapeNet assets once the fix is in....