Lightweight and safe solution to visually detect Air Traffic for "Detect & Avoid" towards BVLOS (EASA/FAA).
Visual safe landing spot detection for package delivering or emergency landing towards BVLOS (EASA/FAA).
Accurate redundant Positioning with vision based SLAM for flying in Indoor-, GPS-denied- or highly populated zones (SORA).
Single Object Tracking under heavy conditions (occlusions) with Geolocalisation for camera Gimbals or automated Flights.
Realtime Multi-Object Detection and Geolocalisation (Persons, Vehicles) on the Ground.
Fully automated Flying in unknown environments for realtime mission planning and highly efficent inspection.
Self-Supervised Training at Large Scale and therefore designed for Domain Adaptation
LIDAR based SLAM (Top) & Semantic Scene Parsing on Point Clouds (Down)
Modern AI technologies, e.g. neural networks, are very powerful due to their fundamental approach of processing massive data in parallel. This statistical property has led to the fact that no AI approach has yet been approved for highly safety-critical systems such as: A – BVLOS Drones, B – Automated Road Vehicles and C – Automated Manned Aircrafts. Since their regulatory requirements like A – SORA, B – ISO26262 / SOTIF or C – DO178 demand high determinism within the process regulation.
At Spleenlab, we have developed a new approach that for the first time allows perceptual AI to be approved in such fields. Our unique technology is based on four key concepts that are incorporated in the development of all our products: