Real multisensor recordings
Real drone flights are recorded with our own multisensor systems: cameras (thermal, RGB, event, NIR, global shutter), audio arrays, radar, LIDAR and RF antennas, all synchronised and calibrated together. The recordings cover different drone models, flight patterns, distances and environments, from open fields to urban areas. Every recording is labelled consistently across all sensors, giving AI teams reliable ground truth taken from real operating conditions.
Built for sensor fusion
All sensors share the same timeline and geometry, so each detection can be matched across modalities. The data is ready to train models that fuse several sensors, to measure how much fusion improves performance compared with single-sensor approaches, and to develop new fusion methods.
Synthetic data & simulation
Realistic simulated scenes add diverse, perfectly labelled data on top of the real recordings. Drone models, distances, trajectories, weather, lighting and backgrounds can all be varied at will. They extend and complement the real data with large volumes of varied examples, generated quickly for any scenario.
What it contains
- Real multisensor recordings, synchronised and fully labelled
- Synthetic data with exact labels for every frame
- Decoys such as birds, to distinguish drones from other elements
What it enables
- Train detection, classification and tracking models
- Develop and benchmark multisensor fusion
- Evaluate models on difficult cases
- Skip costly field campaigns
Who it is for
Computer-vision and AI companies building detection models, counter-drone and security companies, defence and public-safety organisations, research labs and universities.
Early access or a custom dataset? contact@arkiona.ch