FieldSAFE: Dataset for Obstacle Detection in Agriculture

Area of Science:

• Agricultural robotics
• Computer vision
• Sensor fusion

Background:

• Obstacle detection is crucial for autonomous agricultural machinery safety and efficiency.
• Existing datasets may not cover the full range of agricultural scenarios and sensor modalities.

Purpose of the Study:

• To introduce a comprehensive multi-modal dataset for agricultural obstacle detection.
• To provide a valuable resource for developing and evaluating autonomous systems in agriculture.

Main Methods:

• Collected approximately 2 hours of raw sensor data from a tractor-mounted system.
• Utilized multiple sensing modalities: stereo, thermal, web, 360° cameras, LiDAR, and radar.
• Integrated precise localization using fused Inertial Measurement Unit (IMU) and Global Navigation Satellite System (GNSS) data.

Main Results:

• The dataset includes diverse static and moving obstacles such as humans, vehicles, and vegetation.
• All detected obstacles are annotated with ground truth object labels and precise geographic coordinates.
• The data was collected in a grass mowing scenario in Denmark.

Conclusions:

• The presented dataset offers a rich resource for advancing agricultural robotics and autonomous systems.
• It facilitates research in sensor fusion, object recognition, and localization for agricultural applications.