Artifacts in Data Pipeline?

noamgat commented 4 years ago

I'm trying to perform some post-processing on the shapenet meshes for followup research, so I wanted to get the normal pipeline up and running. I am seeing some artifacts being generated in the watertight stage of the process. Attached an example of the 1_scaled and 2_watertight stages for a chair model. Is there something wrong with the setup? Have you experienced this as well?

noamgat commented 4 years ago

I'm having decent success by adding a "cleanup" phase afterwards: After the "produce watertight meshes" phase of build.sh I put this:

echo "Cleaning watertight meshes"
  lsfilter $build_path_c/2_watertight $build_path_c/3_watertight_clean .off | parallel -P 1 --timeout $TIMEOUT \
    meshlabserver -i $build_path_c/2_watertight/{} -o $build_path_c/3_watertight_clean/{} -s dataset_shapenet/clean_mesh.mlx;

And added the following meshlab filter file (put in dataset_shapenet/clean_mesh.mlx):

<!DOCTYPE FilterScript>
<FilterScript>
 <!-- Called "Small component selection" in older versions of meshlab, might need to change filter name -->
 <filter name="Select small disconnected component">
  <Param name="NbFaceRatio" tooltip="This ratio (between 0 and 1) defines the meaning of &lt;i>small&lt;/i> as the threshold ratio between the number of facesof the largest component and the other ones. A larger value will select more components." value="0.01" description="Small component ratio" isxmlparam="0" type="RichFloat"/>
  <Param name="NonClosedOnly" tooltip="" value="false" description="Select only non closed components" isxmlparam="0" type="RichBool"/>
 </filter>
 <filter name="Delete Selected Faces"/>
 <filter name="Remove Zero Area Faces"/>
</FilterScript>

Its not production ready yet, as lsfilter doesn't know to skip duplicate files with this file/folder structure. But I think its a step in the right direction

aprath1 commented 11 months ago

Hi @noamgat ,

Did you manage to get a pipeline which overcomes these artifacts? I mean any other developments besides the above mentioned meshlab cleanup&filter...

autonomousvision / occupancy_networks

Artifacts in Data Pipeline? #41