This release brings new targets and many improvements.

Builds for PyTorch 1.10.1, 1.10.2, 1.11.0, but no longer 1.6.0. Builds for Python 3.10 but no longer 3.6.

New Features

• MultinomialRaysampler and NDCMultinomialRaysampler replace GridRaysampler and NDCGridRaysamplercommit which can sample using the new n_rays input and also bring stratified sampling along rays and direction normalization (alsocommit
• Function join_cameras_as_batchcommit
• Function join_pointclouds_as_batchcommit
• Camera batches can be indexedcommit andcommit
• Meshes.submesh function to take a set of faces from a mesh.commit andcommit
• L1 support for KNN and chamfercommit

Bug fixes

• Rasterizer.to broken without camerascommit
• Joining batches of meshes with TexturesAtlas and TexturesUV broke first input meshes’ texturecommit
• Transform3D.stack was entirely broken.commit which also added typing.
• The function cameras_from_opencv_projection always created on CPU#1021commit
• Batching didn’t work for AmbientLights#1043commit
• Pointclouds.subsample failed on windows#1015commit
• Fit_textured_meshes tutorial now turns off perspective_correct in the final optimization, to avoid nans.commit

Improvements

• Points_normals much faster through use of symeig workaroundcommit
• A warning is now printed to console on rasterizer bin overflowcommit

Small improvements

• MeshRendererWithFragments andmatrix_to_axis_angle made more importable
• FacePointDistance, FacePointDistance and point_mesh_face_distance get min_triangle_area argumentcommit - default also changed incommit
• Fix for small faces in IoU3Dcommit
• Lower the epsilon value in the IoU3Dcommit
• Flexible background color for point compositingcommit andcommit
• In points_to_volumes, the rescaling of features is now optional.commit
• LinearWithRepeat layer clarified and moved inside PyTorch3D from the NeRF projectcommit
• HarmonicEmbedding moved inside pytorch3d from projects/nerf.commit andcommit
• Invalid default values in Meshes.init removedcommit
• Support obj files missing usemtl declarationscommit
• CamerasBase.unproject_points typingcommit
• Fix for Pointclouds.init in certain cases with None in a listcommit
• Consistency validation for sampling_mode in joining TexturesUVcommit
• Doc fix for camera space fix from v0.6.1commit
• Doc fix for renderer picturescommit
• New ndc_grid_sample to handle grid-sampling using our NDC conventionscommit
• Fix scatter_ error in cubify on CPUcommit
• In Transform3D, fix some dtype inconsistenciescommit
• Fixes to make windows builds happiercommit

Internal

• Rename pytorch3d.common.types to pytorch3d.common.datatypescommit
• Cleanup in ply_iocommit
• A major new feature, a framework for new-view synthesis via modeling neural-network based representations, called implicitron, is work in progress.
• We now use usort not isort for sorting imports.
• Future proofing calls to torch.meshgridcommit and symeigcommit alsocommit. Also future proofing tests for a change to autograd behaviour that inplace operations don’t need to return the same objectcommit
• Future proofing for the loss of TH headers in pytorch (needed for PyTorch 1.11)commit andcommit
• Base class for shaders introducedcommit