NVIDIA Omniverse and DRIVE Sim Synthetic Data Generation for AV Training and Validation

We will explore the use of NVIDIA's Omniverse and DRIVE Sim platforms to generate synthetic data to train deep learning networks for autonomous vehicle applications such as object detection, lane keeping, and parking. We will demonstrate using the Omniverse Synthetic Data APIs to extract ground truth data, as well as applying Python scripting to the creation of training scenes, making use of Structured Domain Randomization and generative modeling techniques.