Dataset Creation for Drone Flight Path Generation with LLM
Atori IKEYAMA, K. SATO, Sho Yamauchi, Keiji Suzuki · The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec) · 2024
Recent developments in drone technology have expanded their use in aerial filming. However, capturing footage over large areas remains a challenge, requiring advanced skills and significant labor. This study proposes a straightforward method to automate the process of creating datasets from drone-captured videos. Our approach enables drones to operate based on natural language instructions, focusing on overcoming the previously identified challenge of generating such instructions. By integrating image recognition with Large Language Models (LLM), we provide a practical solution to enhance the automation of drone filming. This method simplifies the dataset creation process, making it more accessible for tasks requiring aerial footage, and offers a step towards more intuitive drone control in various applications.