Simulating Aerial Event-Based Environment: Application to Car Detection

Ismail Amessegher, Hajer Fradi, Clémence Liard, Jean-Philippe Diguet, Panagiotis Papadakis, Matthieu Arzel

Abstract: With the primary goal of enhancing the efficiency of drones for research and rescue missions through the exploitation of neuromorphic sensors and event-based vision, our focus in this work lies in setting up a simulated environment that can be used for synthetic data generation. In particular, we employ Unreal Engine to generate scenes suitable for the case of vehicle perception, followed by a dynamic event-based simulation environment in interaction with the AirSim and v2e software tools. The synthetic event data acquired in this simulated environment is shown to provide a valuable resource for training Artificial Intelligence (AI) systems and more particularly for the task of car detection using YOLOv7.