Start of main content

How to find your place in the 3D world and more with TensorFlow Graphics and PyTorch3D

Language -RU

Recently, computer graphics have been actively used to solve computer vision problems. In this regard, the mastodons of the field of deep learning tools development (Google and Facebook) expand the functionality of their frameworks (TensorFlow and Pytorch), differentiated by graphics layers. As a consequence two libraries TensorFlow Graphics and Pytorch3D were developed, which can be used to solve problems related to 3D data.

In this talk the speaker will answer the questions:

  • how to determine one's position in 3D space from a single photo;
  • which 3D data related tasks can now be solved using Python without resorting to C++;
  • what are the major differences in implementations of graphic layers in frameworks and how this can affect optimization results.