A crucial element in First-Person View (FPV) drone racing is the ability to quickly identify and navigate through a series of gates. Human racers are easily able to identify these gates in a matter of milliseconds. Now, the challenge is for a drone with an embedded computer to do the same. Part of this research overlapped with participation in Lockheed Martin and The Drone Racing League's AlphaPilot AI Drone challenge. This challenge served as valuable experience for the research and provided data sets and test cases to use for training and testing. After a thorough literature review on deep learning, computer vision, and autonomous drone racing, it was concluded that traditional computer vision algorithms by themselves are too unreliable and slow due to weaknesses in lighting and gate overlap [1]. In the same paper, Jung, Hwang, Shin, and Shim provide valuable insight into modifying existing models specifically for gate detection. Another paper on computer vision for autonomous drones provides strong evidence to YOLOv3 as the primary network to utilize for autonomous drone racing [11].
Gate Detection Using Deep Learning
2020-03-01
920113 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Pothole Detection Using Deep Learning
IEEE | 2023
|Driver Drowsiness Detection Using Deep Learning
IEEE | 2023
|Driver Drowsiness Detection using Deep Learning
IEEE | 2022
|Small aircraft detection using deep learning
Emerald Group Publishing | 2021
|Driver Drowsiness Detection Using Deep Learning
IEEE | 2023
|