Image Generation for Efficient Neural Network Training in Autonomous Drone Racing

TY - GEN

T1 - Image Generation for Efficient Neural Network Training in Autonomous Drone Racing

AU - Morales, Theo

AU - Sarabakha, Andriy

AU - Kayacan, Erdal

PY - 2020/7

Y1 - 2020/7

N2 - Drone racing is a recreational sport in which the goal is to pass through a sequence of gates in a minimum amount of time, while avoiding collisions. In autonomous drone racing, one must accomplish this task by flying fully autonomously in an unknown environment by relying only on computer vision methods for detecting the target gates. Due to the challenges such as background objects and varying lighting conditions, traditional object detection algorithms based on colour or geometry tend to fail. Convolutional neural networks offer impressive advances in computer vision, but require an immense amount of data to learn. Collecting this data is a tedious process because the drone has to be flown manually, and the data collected can suffer from sensor failures. In this work, a semi-synthetic dataset generation method is proposed, using a combination of real background images and randomised 3D renders of the gates, to provide a limitless amount of training samples that do not suffer from those drawbacks. Using the detection results, a line-of-sight guidance algorithm is used to cross the gates. In several experimental real-time tests, the proposed framework successfully demonstrates fast and reliable detection and navigation.

AB - Drone racing is a recreational sport in which the goal is to pass through a sequence of gates in a minimum amount of time, while avoiding collisions. In autonomous drone racing, one must accomplish this task by flying fully autonomously in an unknown environment by relying only on computer vision methods for detecting the target gates. Due to the challenges such as background objects and varying lighting conditions, traditional object detection algorithms based on colour or geometry tend to fail. Convolutional neural networks offer impressive advances in computer vision, but require an immense amount of data to learn. Collecting this data is a tedious process because the drone has to be flown manually, and the data collected can suffer from sensor failures. In this work, a semi-synthetic dataset generation method is proposed, using a combination of real background images and randomised 3D renders of the gates, to provide a limitless amount of training samples that do not suffer from those drawbacks. Using the detection results, a line-of-sight guidance algorithm is used to cross the gates. In several experimental real-time tests, the proposed framework successfully demonstrates fast and reliable detection and navigation.

KW - convolutional neural networks

KW - deep learning

KW - drone racing

KW - semi-synthetic images generation

KW - unmanned aerial vehicles

UR - http://www.scopus.com/inward/record.url?scp=85093827577&partnerID=8YFLogxK

U2 - 10.1109/IJCNN48605.2020.9206943

DO - 10.1109/IJCNN48605.2020.9206943

M3 - Article in proceedings

AN - SCOPUS:85093827577

BT - 2020 International Joint Conference on Neural Networks (IJCNN)

PB - IEEE

T2 - 2020 International Joint Conference on Neural Networks, IJCNN 2020

Y2 - 19 July 2020 through 24 July 2020

ER -