Product defects are costly for businesses and can lead to financial and consumer health issues. The current dataset contains 168 image files with a total of 1859 jar lids, on average 11 per image. Categories include intact (962) versus damaged jar lids (897). Type of damages are lid deformations, holes and scratches. An annotation file with bounding boxes is added (jarlids_annots.csv).

The data has train, validation and unseen test data:

• p1-p124 are train data,
• p200-p238 are validation data,
• t1-t5 are test data.