SEATTLE, Wash. — February 27, 2018 —
Mighty AI and Mcity are teaming up to provide a new vehicle and pedestrian detection training dataset exclusively for Mcity members to train their machine learning models. Mighty AI also announced today it will join Mcity’s affiliate member program, with Mighty AI investing $150,000 in the public-private R&D initiative over the next three years.
The training dataset is a first-of-its-kind offering for Mcity members and is designed to aid in furthering Mcity’s mission to accelerate the development and implementation of connected and automated vehicles. Mighty AI will deliver the training dataset in the next few months. It will include more than one million labeled objects across thousands of images.
“To produce the high quality research results we’ll need to ensure safety of automated vehicles, we must first start with high quality data,” said Mcity Lab Director Greg McGuire, who oversees the Mcity Test Facility for connected, automated and driverless vehicles. “Mighty AI provides that and more with their data management platform. This is a tremendous addition to the more than two decades of existing automated vehicle data Mcity and other U-M units have collected.”
Test vehicles in Mcity and the greater Ann Arbor area have collected a large amount of raw sensor data of a broad range of driving scenes. This includes real imaging of different vehicle types (passenger, emergency, semi-trucks, construction, etc.) and pedestrian scenarios (crossing the street, waiting at a bus stop, pushing a stroller down a sidewalk, etc.). Mighty AI will leverage its platform—which pairs machine learning with human intelligence to generate ground truth datasets for vehicle perception systems—to ingest this raw data and generate a fully labeled dataset that Mcity members can use to train their machine learning models to detect vehicles and pedestrians both in Mcity’s testing facility and beyond.
“Companies developing autonomous vehicle technology can have the best AI researchers in world and cutting-edge models, but if they don’t have the right training data to make their systems a reality, they won’t get far. As a leader in R&D for connected and autonomous vehicles, Mcity understands how critical high quality training data is to making autonomous driving models safe,” says Mighty AI Founder and CEO Daryn Nakhuda.
Together, Mighty AI and Mcity will help Mcity’s nearly 60 members
—which represent a broad range of automotive manufacturers and suppliers, chip and hardware makers, insurance providers, and more—advance their machine learning models. Mighty AI’s platform for managing ground truth datasets for computer vision and machine learning models includes a community of more than 400,000 annotators globally who specialize in labeling vehicle sensor data. The resources Mighty AI brings to bear are an ideal pairing with Mcity, which is the only advanced mobility R&D center that combines early-stage research; testing in the state-of-the-art, real-world environment of the Mcity Test Facility; and on-road vehicle deployments to further prove new technologies.
About Mighty AI
Mighty AI is a data management platform for autonomous vehicle perception data. We provide all the software, expertise, and expert annotators needed to create ground truth datasets for vehicle perception models. We convert raw, unlabeled data into useful high-quality data, pairing machine learning with human intelligence to revolutionize how companies manage perception data and create ground truth data to train and validate their models. Visit www.mty.ai
to learn more, and follow us at @mighty_ai