Ford Releasing Comprehensive Dataset to Promote Autonomous R&D

May 27, 2020

Paul Seredynski

Ford is releasing a comprehensive autonomous vehicle (AV) dataset to the academic and research community to help spur innovation in the field. The package includes data from multiple self-driving research vehicles collected over a span of one year, part of Ford’s advanced R&D efforts, but separate from the work it’s doing with Argo AI to develop production-ready AV systems. The high-quality dataset can assist in engineering software to properly teach self-driving vehicles how to analyze their environments.

The dataset includes lidar and camera sensor data, GPS and trajectory information, as well as unique elements such as multi-vehicle data and 3D point cloud and ground reflectivity maps. A plug-in is also available to visualize the data, which is offered in the popular ROS format. According to Ford’s Tony Lockwood, autonomous vehicle manager, virtual driver systems, “There’s no better way of promoting research and development than ensuring the academic community has the data it needs to create effective self-driving vehicle algorithms.”

In addition to 3D point cloud maps, access is also being provided to high-resolution 3D ground plane reflectivity maps, which can provide a more comprehensive understanding of what AVs “see” in the world around them. Spanning an entire year, the dataset incorporates seasonal and weather variations for driving environments in the Detroit metro area, including freeways, dense urban areas, construction zones and pedestrian activity.

The high-resolution, time-stamped dataset was generated by Ford engineers on vehicles with four lidar and seven cameras, and features localization and ground truth data to help researchers correlate the accuracy of their algorithms. Provided via Ford’s collaboration with the Amazon open data program, information about accessing the data package is available at avdata.ford.com .