The Synthetic Data Application in the UAV Recognition Systems Development

Abstract

The increasing popularity and accessibility of un-manned aerial vehicles (UAVs) presents both opportunities and challenges. On the one hand, UAVs has a wide range of civilian, industrial, and military applications. On the other hand, the popularity of UAVs can lead to illegal or dangerous usage. Thus, the development of UAV recognition systems is crucial for ensuring safety and security. However, collecting and labeling large amounts of real-world data for training these systems can be time-consuming and labor-intensive.In this study, we propose a methodology, which can help to accelerate the development of new UAV recognition systems. This work demonstrates the effectiveness of training a neural network using a combination of real-world and synthetic data that can achieve similar performance to a network trained on real-world data only.