Enhancing Road Infrastructure Maintenance Using Deep Learning Approach

The physical gathering of info on road damage can be time-consuming and dangerous to people's safety, yet preserving the transportation network is essential for providing secure and lasting transport systems. To address these challenges, this paper presents a novel approach that leverages Unmanned Aerial Vehicles (UAVs) and deep learning approaches can greatly improve the effectiveness and precision of road damage identification. The proposed approach employs the YOLOv8 algorithm, cutting-edge object detection and localization model, for detecting road damage in UAV images. The YOLOv8 model offers improved accuracy compared to previous versions, namely YOLOv4, YOLOv5, and YOLOv7, by using UAVs equipped with high-resolution cameras, info on road damage can be captured efficiently and remotely. On a synthesis of the Indian roadway data and the RDD2022 dataset, this method was trained and tested. The trained YOLOv8 model was used to identify and locate different types of road degradation such cracks, potholes, and surface degradation. The combination of UAV imagery and deep learning algorithms has shown great potential to detect automated road deterioration. This research highlights the advantages of using YOLOv8 for improved accuracy in identifying road damages. This work lays the path for further developments in the field, enabling safer transport systems and more effective road maintenance.