NVIDIA Creates AI Training Environment With Synthetic Data

NVIDIA has developed a synthetic environment that deploys virtual sensor data to help train driverless cars, C4ISRnet reported Saturday.

Margaret Amori, a business development lead at NVIDIA’s federal arm, said at an Association of the U.S. Army symposium that the technology is designed to incorporate realistic and less ideal environments to adequately train self-driving cars without the hassle of physically driving millions of miles.

“It’s very physics based and very realistic and accurate,” she said. “So you can literally move the sun around, you can change it from summer to winter with a click of a button and create all sorts of conditions like sun glare, which is very problematic for our sensors.”

Frank Schirrmeister, a senior group director at Cadence Design Systems, noted that synthetic data can help create and automate a menu of driving information to establish a virtual Army or battlefield scenario.

According to NVIDIA, developers can now leverage an even mix of synthetic and field data to train artificial intelligence technologies and support navigation activities.