Paper Title
DEFECT DETECTION IN PRINTED CIRCUIT BOARDS USING DEEP LEARNING APPROACH

Abstract
Abstract - In this paper, a CNN-based PCB defect detection technique is presented for both assembled and bare PCBs. Even though a lot of work has already been done in the area of defect detection, not much work has been done to develop a single approach that could identify defects in both kinds of PCBs.Thus, a CNN-based model called YOLOv5 small has been used to detect six different types of defects in bare PCBs, as well as two components, ICs and capacitors, which, if absent or improperly positioned, can result in a defective PCB. Finally, in bare PCBs, six defects were detected with a significant mAP of 93.6%, while in assembled PCBs, two components were detected with a significant mAP of 96.1% @IOU using small version of YOLOv5. Keywords - Printed Circuit Boards (PCBs), Convolutional Neural Network (CNN), YOLOv5, Defect detection.