Few-shot Learning for Bagel Defect Detection

Nowadays, the request for quality is increasing more than before, but defective products are found and cannot be prevented during the manufacturing process. Therefore, the defective products are abandoned. However, since each of the defects is not caused by the same process, which needs the manufacturing process to be optimized to reduce the occurrence of defective products. Therefore, manufacturers require an inspection process that provides immediate results. Therefore, using the Prototypical network using with few-shot learning on the bagel dataset, the effect of surface defect detection on 3D objects is proposed with its optimized accuracy in this study.