Autonomous driving relies on a huge volume of real-world data to be labeled to high precision. Alternative solutions seek to exploit driving simulators that can generate large amounts of labeled data with a plethora of content variations. However, the domain gap between the synthetic and real data remains, raising the following important question: What are the best way to utilize a self-driving simulator for perception tasks? In this work, we build on top of recent advances in domain-adaptation theory, and from this perspective, propose ways to minimize the reality gap. We primarily focus on the use of labels in the synthetic domain alone. Our approach introduces both a principled way to learn neural-invariant representations and a theoretically inspired view on how to sample the data from the simulator. Our method is easy to implement in practice as it is agnostic of the network architecture and the choice of the simulator. We showcase our approach on the bird's-eye-view vehicle segmentation task with multi-sensor data (cameras, lidar) using an open-source simulator (CARLA), and evaluate the entire framework on a real-world dataset (nuScenes). Last but not least, we show what types of variations (e.g.

Autonomous driving relies on a huge volume of real-world data to be labeled to high precision. Alternative solutions seek to exploit driving simulators that can generate large amounts of labeled data with aplethora of content variations. However, the domain gap between the synthetic and real data remains, raising the following important question:What arethe best way toutilize aself-driving simulatorforperceptiontasks?

Autonomous driving relies on a huge volume of real-world data to be labeled to high precision. Alternative solutions seek to exploit driving simulators that can generate large amounts of labeled data with a plethora of content variations. However, the domain gap between the synthetic and real data remains, raising the following important question: What are the best ways to utilize a self-driving simulator for perception tasks? In this work, we build on top of recent advances in domain-adaptation theory, and from this perspective, propose ways to minimize the reality gap. We primarily focus on the use of labels in the synthetic domain alone. Our approach introduces both a principled way to learn neural-invariant representations and a theoretically inspired view on how to sample the data from the simulator. Our method is easy to implement in practice as it is agnostic of the network architecture and the choice of the simulator. We showcase our approach on the bird's-eye-view vehicle segmentation task with multi-sensor data (cameras, lidar) using an open-source simulator (CARLA), and evaluate the entire framework on a real-world dataset (nuScenes). Last but not least, we show what types of variations (e.g.

Nvidia Corp. has advanced deep learning techniques, but now it's looking to take AI technology into new areas: Putting self-driving cars into virtual reality instead of our roads, and setting its sights on Hollywood and hospitals. Over the past few years, Nvidia has made inroads into equipping cars with the computer hardware that gives them self-driving capability. That move has become so crucial that Nvidia NVDA, -7.76% shares fell more than 6% in recent trading as the company kicked off its GPU Technology Conference in San Jose, Calif., after it confirmed that it is suspending real-world testing following a recent fatality in Arizona in one of Uber Technologies Inc.'s self-driving cars. In his keynote address Tuesday morning, Chief Executive Jensen Huang did not mention the halt, but did show off a potential solution to the problem of testing self-driving automobiles on public roads. Huang showed off a simulator that can allow companies to test their self-driving systems in a virtual environment, providing opportunity to drive billions of miles in a year without endangering pedestrians.