Coating Defect Detection Dataset

This dataset includes 2697 RGB images of Wind Tower Painting parts, along with bounding box annotations (YOLO format) for defect localization and classification. In 'class_labels.txt' a legend of the three labels of the defect classes is given. In 'metadata.csv' each image filename is related to the acquisition or section setup code. It also indicates the main label (in some images there is more than one defect), if a cold or warm white balance and high or low surface contrast are captured.