PyTorch3D: An open-source 3D deep learning library

Understanding it 3D plays an important role in promoting their capabilities AI systems. It Facebook AI Research (FAIR) presented it yesterday PyTorch3D, a library of unique features that enables researchers and developers to combine it deep learning with 3D data.

In addition, Facebook is now doing Mesh R-CNN, open-source. Mesh R-CNN is one system created last year and can convert 2D shapes into 3D objects. According to FAIR engineer Nikhila Ravi, the researchers created the new PyTorch3D by taking data from Mesh R-CNN and the researchers' recent work on 3D projects.

Dealing with 3D is very important for rendering XNUMXD objects or scenes that appear in mixed or virtual reality conditions. It can also be used to deal with many of them the challenges it faces AI. For example, it could help stand-alone vehicles understand the location of nearby objects.

PyTorch3D comes with many new functions 3D, which are not available in any of the existing 3D libraries.

“With PyTorch3D, researchers can introduce all the new features and use them with existing ones system deep learning in PyTorch. This significantly reduces working time for 3D design research, which requires a lot of work and experience. We want to try to reduce this time, "Ravi said.

The researchers believe that PyTorch3D will make it research more fast and flexible.

The release of PyTorch3D comes after PyRobot robotics framework, which was introduced last year.