CarDD: A New Dataset for Vision-Based Car Damage Detection

Automatic car damage detection has attracted significant attention in the insurance business. However, due to lack of high-quality and publicly available datasets, we can hardly learn a feasible model for detection. To this end, contribute with Car Damage Detection (CarDD), first public large-scale dataset designed vision-based segmentation. Our CarDD contains 4,000 high-resolution images over 9,000 well-annotated instances six categories. We detail image collection, selection, annotation processes, present statistical analysis. Furthermore, conduct extensive experiments on state-of-the-art deep methods different tasks provide comprehensive analyses highlight specialty source code are at https://cardd-ustc.github.io .