Asphalt Cracked and Uncracked Image Dataset

This dataset consists of 2000 cracked and uncracked images of asphalt pavement with an image resolution of 1440 x 1440, more than 8000 cracked and uncracked images of resolution 360x360 and more than 3000 cracked and uncracked images of resolution 720x720. The datasets of images with resolution 360x360 and 720x720 were created by splitting the original 1440 x 1440 dataset and picking the cracked and uncracked sub images. Original dataset was created using the same scale by ensuring the same image resolution of the camera and same working distance and field of view by mounting camera at same height , and hence, crack widths and lengths can be compared.
In order to collect the dataset, a vehicle-mounted camera enclosure system was used. A wooden chamber was constructed to enclose the camera with a high resolution of 1440x1440 and a frame rate of 30 fps. This wooden chamber was constructed from plywood sheets that were glued and screwed together, as well as wooden beams and dowels for reinforcement. The wooden dowels were used to secure the system to a commercial bike carrier. A Perspex screen was attached to the bottom surface of the chamber to prevent the camera from falling during the operation. The camera has a clear and unobstructed view downwards due to the transparency of the Perspex screen. This structure was then suspended from a commercially available bicycle carrier and installed on the towbar of a car which enabled filming while driving along the target road.
Please refer to the published journal paper in Construction Building Materials for more information about this dataset and applications. This dataset can be used to train deep learning algorithms and develop algorithms to calculate the crack widths of asphalt pavements.