Annotated image record “ARC-D” on asphalt damage on roads

The data set contains 19,735 images for the classification of asphalt damage sites into four categories and background images of the road periphery. The dataset is divided into training and validation data. The data classes are specified by the names of the folders. The training data includes the classes: — Longitudinal crack (longitudinal crack) with 2,883 images — Transversal crack (cross crack) with 2,877 images — Net crack with 5,386 images — Pothole with 1,301 images — Background with 7.288 images The validation data has been manually cleaned and is available with 100 images per class each. All images are available in.jpg format. The data set “ARC-D” is GDPR compliant, as all faces and number plates have been rendered unrecognizable by a visual inspection.