This PyTorch Library ‘Kaolin’ is Accelerating 3D Deep Learning Research

NVIDIA released a PyTorch library ‘Kaolin’, which in few steps, moves 3D models into neural networks. Kaolin helps in easy implementation of 3D modules for use in deep learning models.

Kaolin is developed with advanced functionalities to load and preprocess multiple 3D datasets and functions. The advanced nature of Kaolin helps in preparing 3D models for deep learning from 300 lines of codes to five lines.

In terms of applications, Kaolin can help researches in virtual and augmented reality, robotics, medical imaging, etc.

Github: https://github.com/NVIDIAGameWorks/kaolin

Paper: https://arxiv.org/abs/1911.05063

Blog: https://news.developer.nvidia.com/kaolin-library-research-3d/

Installation

First create a virtual environment. In this example, we show how to create a conda virtual environment for installing kaolin.

$ conda create --name kaolin python=3.6
$ conda activate kaolin

Now, install the dependencies (numpy and torch). Note that the setup file does not automatically install these dependencies.

conda install numpy

Install PyTorch, by following instructions from https://pytorch.org/

Now, you can install the library. From the root directory of this repo (i.e., the directory containing this README file), run

$ python setup.py install

During installation, the packman package manager will download the nv-usd package to ~/packman-repo/ containing the necessary packages for reading and writing Universal Scene Description (USD) files.

Verify installation

>>> import kaolin as kal
>>> print(kal.__version__)

Installation Details: https://github.com/NVIDIAGameWorks/kaolin