Litcius/Paper detail

Canned Food Surface Defect Classification Using YOLOv4

Aljohn G. Asenci, Joshua Emmanuel D.C. Bulawan, Dionis A. Padilla

202212 citationsDOI

Abstract

According to Association of Food and Drug Officials (AFDO), the canned food's defect can be classified into Class 1 – Critical Defect, Class 2 – Major Defect, and Class 3 Minor Defect depending on the severity of such defect in damaging the can's hermetic seal. The researchers propose the utilization of YOLOv4 algorithm to classify the defects as per the AFDO standard and implement such model in a multi camera setup connected to a Raspberry Pi 4. The system will capture images from the left-side view, top view, and right-side view of the can and feed the image to the model. The resulting classification of the three images will then be stitched together to be further classified into one final classification. The system performed well, having an overall accuracy of 93%.

Topics & Concepts

Raspberry piComputer scienceClass (philosophy)Artificial intelligenceContextual image classificationSeal (emblem)Pattern recognition (psychology)Computer visionImage (mathematics)Computer securityInternet of ThingsGeographyArchaeologyIndustrial Vision Systems and Defect Detection
Canned Food Surface Defect Classification Using YOLOv4 | Litcius