Autonomous vehicles equipped with sensors collect vast amounts of data. FL allows these vehicles to collaboratively train models for improving self-driving algorithms. Autonomous driving (AD) technology can dramatically reduce the number of car accidents, but what is needed is a great scale of sensor data collected from real-world driving under various driving conditions. One real-world example is Zenseact Open Dataset (ZOD).

Zenseact decided to share public data to collaborate and join forces to improve rode safety. This dataset is a large multi-modal AD dataset. It was collected over a 2-year period in 14 different European countries using a fleet of vehicles equipped with a full sensor suite. To find out more information about ZOD, visit their website or refer to their article .

ZOD provides a comprehensive set of annotations for different purposes, such as object detection, lane marking identification, and road condition assessment:

• Static and Dynamic Objects, 2D and 3D Bounding Boxes
  This could help detect and classify objects like vehicles, pedestrians, and other obstacles in the scene. It is essential for the development of object detection and recognition algorithms in autonomous vehicles.
• Lane Markings, Instance and Semantic Segmentation of Lane Markings and Road Paintings
  These annotations could be used to identify and segment lane markings on the road, which is crucial for lane-keeping assist systems, lane departure warnings, and autonomous lane changing.
• Ego Road: Semantic Segmentation of the Ego-Road
  Identifying the drivable area (ego-road) allows the vehicle to understand which parts of the road are safe for driving. This helps in path planning and maintaining the vehicle within safe boundaries.
• Traffic Sign Recognition
  Recognizing and understanding traffic signs enable vehicles to obey traffic rules, such as speed limits, stops, and yields. This enhances road safety and compliance with traffic regulations.
• Road Condition Classification
  Assessing road conditions allows vehicles to adjust their driving behavior according to the surface state. For example, a wet or snowy road might require slower speeds and more cautious maneuvers to prevent skidding.