The Drive&Act dataset is a state of the art multi modal benchmark for action recognition in automated vehicles. It offers following key features:

  • 12h of video data in 29 long sequences
  • Calibrated multi view camera system with 5 views
  • Multi modal videos: NIR, Depth and Color data
  • Markerless motion capture: 3D Body Pose and Head Pose
  • Model of the static interior of the car
  • 83 manually annotated hierarchical activity labels:
    • Level 1: Long running tasks (12)
    • Level 2: Semantic actions (34)
    • Level 3: Object Interaction tripplets [action|object|location] (6|17|14)


Copyright Fraunhofer IOSB.

Usage for research only.

If you have any questions regarding the dataset please contact: Manuel Martin and Alina Roitberg.

Location Video
Center mirror NIR
A-Column co-driver NIR Depth RGB Kinect IR
A-Column driver NIR
Ceiling NIR
Steering wheel NIR
Annotations Activities 3D Body Pose Interior


If you use the dataset please cite the following publication.

Drive&Act: A Multi-Modal Dataset for Fine-Grained Driver Behavior Recognition in Autonomous Vehicles

We introduce the novel domain-specific Drive&Act benchmark for fine-grained categorization of driver behavior. Our dataset features …


Access restrictions removed

We removed the login requirements. Data is now available without asking for permission.

Project page setup for ICCV

Configured the webpage. Providing the data used in the ICCV publication.
Coming soon: head pose, continuous labels and a more detailed interior model


Fraunhofer IOSB

Karlsruhe Institute of Technology (KIT)