Attain

An inclusive pavement distress dataset containing 2,293 images of 19,761 distress-type instances taken with smartphone cameras mounted on a vehicle's front and rear windshields. This dataset comprises images of ten different pavement distress types, including longitudinal and transverse cracks (linear cracks), alligator cracks, block cracks, weathering, lane/shoulder drop-off, raveling, patching and utility cuts, manholes, faded markings, and potholes. The Attain dataset's images have been manually annotated for object detection and classification purposes and can be used to train machine learning and deep learning models to detect and classify pavement distresses.