Deep-learning-based detection of missing road lane markings using YOLOv5 algorithm

Abstract

Road lane markings are critical for ensuring road user safety. To improve their safety, there are even different types of road lane markings, such as single solid lines, double solid lines, dashed single line etc. Their colors generally are white and yellow. This road lane markings mainly used to provide guidance and information for road user to comply with the rule of the road. Unfortunately, these markings get worn out with time and may even disappear. In order to prevent this from happening, regular inspection and maintenance need to be conducted. Manual inspection is tedious, slow, and prone to human errors. With the recent technological advancement, especially in machine vision and artificial intelligence, automated or semi-automated missing road lane marking detection systems can potentially be developed. In this work, preliminary study of the implementation of one of the latest deep learning algorithms, i.e. YOLOv5, has been carried out in the detection and classification of missing road lane markings. This paper shows the preliminary results which look promising as the mean Average Precision (mAP@0.5) reaches 0.995.