Road Damage Detection and Classification Using Deep Neural Networks (YOLOv4) with Smartphone Images

Abstract

Research on damage detection of road surfaces using image processing techniques has been actively conducted achieving considerably high detection accuracies. However, many studies only focus on the detection of the presence or absence of damage. However, in a real-world scenario, when the road managers from a governing body needs to repair such damage, they need to know the type of damage clearly to take effective action. In addition, in many of these previous studies, the researchers acquire their own data using different methods. Hence, there was no uniform road damage data set available openly, leading to the absence of a benchmark for road damage detection. For the first time, road damage detection and classification challenge (one of the IEEE Big-data Cup Challenge) was held in Seattle provided such a big dataset for pavement damage detection and classification. In this study TensorFlow implementation of tiny-YOLOv2, Dark-net Neural Networks YOLOv3, tiny-YOLOv3 and YOLOv4 were used to train a road damage detection model with the data set provided by the IEEE Big-data Cup Challenge, and results were compared in the term of the accuracy and runtime speed with other similar studies using different models