No abstract

No abstract

Systems for automatic inspection of product quality are in high demand. However, their prevalence is limited by complex development and great expenses. Since inspection systems must be engineered to specific products and environments, such systems are generally only viable with high volume product series. Inspired by human visual inspection of highly reflective brushed aluminium objects we capture images across multiple viewpoints. We employ a spatio-temporal weighting of defects, where defects that occur consistently across viewpoints are considered more severe, and compare to the confidence scores produced by an off-the-shelf object detector (YOLOv5). Our results show the challenges with training object detectors on a realistic low-volume dataset of reflective brushed surfaces. Despite the poor detection performance and difficulty in distinguishing between design and defects, our method proves to classify our small test set with an area under the precision-recall curve of 66.5%.

No abstract

No abstract