We can create digital twins of real sites, such as mines, to simulate in.
Then use WorldGenerator to edit the digital twin to create unique scenarios for validation and verification of autonomous systems.
Synthetic data is an excellent tool to increase the speed of autonomous vehicle development. We believe it’s optimally used in compliment to logged data, to provide a high quantity, high quality training dataset.
Logged data is dependent on the weather at the time of collection. No need to wait with simulators and synthetic worlds.
Capturing ‘rare events’ during manual logging of data is, well, rare. Synthetic environments can simulate these rare events.
We’ve removed one of the largest time sinks in training a perception system. This means no more frame-by-frame annotation processes.
Dangerous or high-risk scenarios are difficult to train due to potential harm of life or risk of destroying engineering components. The safest way to train the perception system is in a virtual world.
© Repli5 AB. All rights reserved.