The University of California, Berkeley (UC Berkeley) has made available a massive dataset used by engineers in the development of driverless auto technologies, featuring more than 100,000 high-definition video sequences that represent different driving situations.
The downloadable BDD100K dataset is part of the university's DeepDrive project, and it also contains global-positioning system locations, inertial measurement unit data, and timestamps across 1,100 hours.
Such datasets are needed to train systems on how to deal with different environments and driving conditions.
The BDD100K also has two-dimensional bounding boxes that have annotated approximately 100,000 images with notable objects such as traffic signs, pedestrians, bicycles, other vehicles, trains, and traffic lights.
"To achieve rich annotation at scale, we found that existing tooling was insufficient, and therefore develop novel schemes to annotate driving data more efficiently and flexibly than previous methods," the UC Berkeley team says.
View Full Article
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA
No entries found