ABSTRACT
Road distress evaluation plays a critical role in pavement maintenance systems. Traditional methods of collecting pavement distress data involve manually identifying distress types and severity levels, followed by data analysis to assess pavement conditions. These methods are time-consuming and prone to errors, highlighting the need for automation to ensure timely and accurate detection and classification of pavement distress. This study explores the use of deep learning neural network ResNet50 to classify two common asphalt pavement distresses, raveling (surface distress) and alligator cracks (structural distress), and to assess their severity levels (high, medium, low). Images were collected using mobile devices from the provincial highways of Pakistan. This research aims to optimize the model’s performance by adjusting how well it performs on validation data, ensuring thorough training of distress types and severity models. The research paves the way for implementing an automated system that could significantly enhance the efficiency of pavement maintenance.
