Home // ICAS 2020, The Sixteenth International Conference on Autonomic and Autonomous Systems // View article

Reference Detection for Off-road Self-Driving Vehicles Using Deep Learning

Authors:
Marcelo Pederiva
Ely de Paiva

Keywords: YOLO; Faster RCNN; SSD; Object Detection; Autonomous Vehicles.

Abstract:
This paper proposes the application of deep neural network models to detect references in off-road driving for autonomous vehicles. Due to the absence of traffic signs in non-urban areas, the work searched for a low-cost sensory-based solution for autonomous localization in this environment. Given the advancement of Machine Learning techniques, we used Object Detection algorithms to solve the localization problem. For this reason, we trained three existing object detection models (Fast YOLOv2, SSD300 and Faster R-CNN) to detect a reference at the road boundary. The project analyzed these three architectures performance after training with a small dataset (around 300 images), regarding the detection distance, the number of detection and image processing time. Through two experiments, one in the same environment as the training step and another with a different background, we evaluate the pros and cons of each model and the possible application scenario for each one in autonomous cars.

Pages: 97 to 102

Copyright: Copyright (c) IARIA, 2020

Publication date: September 27, 2020

Published in: conference

ISSN: 2308-3913

ISBN: 978-1-61208-787-0

Location: Lisbon, Portugal

Dates: from September 27, 2020 to October 1, 2020