Skip to content

v0.9.1

Compare
Choose a tag to compare
@Caenorst Caenorst released this 03 Oct 22:30
· 186 commits to master since this release
a9b007a

Changelogs:

Summary

The latest Kaolin release includes a new representation, structured point clouds, an octree-based acceleration data structure, with highly efficient convolution and ray tracing capabilities. This representation is useful for neural implicit representations, popular in 3D DL research today, and beyond, and powers the latest version of NeuralLOD training.

The release is also coming with extended support for 3d dataset like ModelNet / ShapeNet / SHREC, new utility functions to improve usability and speedups on import / export of USD used in checkpoints.
In this version, we added a lightweight visualizer Dash3D for quickly visualizing from a low-end config such as a laptop.

New Additions

Features

  • added pointclouds_to_voxelgrids (#361)
  • added support for non-fully connected mesh for uniform_laplacian (#380)
  • added mask_IoU on rendered images (#372)
  • added support for camera transform matrix (instead of just rotation / translation) (#372)
  • support for SHREC (#375)
  • support for colors in exporting point clouds in USD (#400)
  • support for UsdGeomPoints (#400)
  • support for .off (#367)
  • support for ModelNet (#382, #410)
  • added utility function for loading synthetic data from OV app (#372)
  • added material support for ShapeNet (#388)
  • added version support for ShapeNet (#399)

Optimizations

  • accelerated USD import: ~10-5X (#421)
  • accelerated USD export: ~8-4X for exporting and timelapse (#422)
  • accelerated backward of index_by_face_vertices (#419)

Bug fix:

  • fixing a bug on texture_mapping removing when UVs are out-of-bounds.
    fix some issues with ShapeNet and support for bad models (#391, #411)

Misc:

  • Allow users to install Pytorch version out of official support (#390)

Contributors:

  • Clement Fuji Tsang
  • Masha Shugrina
  • Charles Loop
  • Towaki Takikawa
  • Jiehan Wang
  • Michael Li
  • @AndresCasado
  • @mjd3