Analysis of YOLO V3 for Multiple Defects Detection in PCB

The work proposes an analysis of YOLO v3 algorithm for PCB defect detection during the PCB manufacturing phase. Utilizing the YOLO v3 architecture, PCB defect detection benefits from its grid-based approach, anchor boxes, multi-scale operation, and efficient non-maximum suppression. The grid division enables systematic object prediction, ensuring thorough coverage for fault detection. Anchor boxes aid in precise component identification and localization, accommodating diverse component sizes. Multi-scale operation enhances detection across varying object sizes, while efficient non-maximum suppression minimizes redundant predictions, reducing false positives. Experimental results demonstrate, YOLO v3's features make it an effective tool for accurate detection of single and multiple PCB faults with an overall mean average precision pf 90%.