Skip to main content
SHARE
Project

ORNL Database to Enable Face Characterization in Driving Studies (DEFADS)

Topic:
ORNL Database to Enable Face Characterization in Driving Studies (DEFADS)

The ORNL Database to Enable Face characterization in Driving Studies (DEFADS) is a dataset created by Oak Ridge National Laboratory to help evaluate algorithms used in an analysis of automobile driver behavior. The sample was developed as part of the U.S. Department of Transportation’s Federal Highway Administration Exploratory Advanced Research Program efforts seeking to understand the role of driver performance and behavior on traffic safety.

DEFADS is the second dataset created by ORNL to emulate aspects of the Second Strategic Highway Research Project’s (SHRP2) Naturalistic Driving Study (NDS). The NDS study collected data from more than 3,000 drivers in personal vehicles between 2010 and 2013, which were then analyzed using algorithms to assess how drivers interact with and adapt to their vehicle, traffic environment, roadway conditions, and traffic control devices, including assessment of collision risk.

Both DEFADS and the Oak Ridge Naturalistic Driving Study (ONDSS) utilized a SHRP2 data acquisition system borrowed from Virginia Tech Transportation Institute.  The SHRP2 face camera is positioned behind the rear-view mirror similar to the real SHRP2 study.  DEFADS includes two additional camera views, one with a higher resolution color camera at the same position as the SHRP2 face camera and another in front of the driver.  The data set contains over 70 subjects and features simulations of drivers in various prescribed motions and poses that might be seen during vehicle operation. The activities were performed in a parked vehicle with head poses and other motions captured by the cameras.  Some processing of the data has been performed as well for face characterization and is included in the dataset. 

The data collection was supported by the Exploratory Advanced Research Program of the Federal Highway Administration, with assistance from the Virginia Tech Transportation Institute.

For the full dataset send a request to karnowskitp@ornl.gov , aykacdb@ornl.gov, and torkelsonlc@ornl.gov

More details on the dataset here.