Attain: Inclusive annotated pavement distress types and severity dataset

Pavement distress detection plays a crucial role in pavement management and rehabilitation (M&R) by providing essential data for maintenance decision-making. The Attain dataset is introduced to facilitate the development of machine learning and deep learning models for pavement distress classification and object detection. The dataset consists of 2293 images collected using three smartphones mounted on vehicle windshields. All photos are manually annotated by pavement engineer experts. The images capture a variety of pavement conditions under different lighting and traffic scenarios, making the dataset diverse and suitable for training robust models. The dataset encompasses ten distinct pavement distress categories including alligator cracks, block cracks, longitudinal and transverse (linear) cracks, faded marking, lane/shoulder drop-off, patch and utility cut, potholes, manhole, weathering, and raveling at two levels of severity i.e., low, medium, and high. In total, 19,761 distress-type instances were detected and annotated with distress type and severity. The core competency of the Attain dataset is the wide range of pavement distress types included and the severity levels of each distress provided. The use of readily available smartphone cameras significantly reduces the cost of data collection while maintaining high-quality image acquisition. The Attain dataset aims to support researchers and practitioners in developing automated pavement distress detection systems to enhance the efficiency and accuracy of pavement maintenance processes.