Enhancing EfficientNet-YOLOv4 for Integrated Circuit Detection on Printed Circuit Board (PCB)

Abstract

Ensuring printed circuit board (PCB) quality and functionality during manufacturing demands precise automated visual inspection. Detecting integrated circuits (ICs) on PCBs poses a significant challenge due to diverse component sizes, types, and intricate board markings, complicating accurate object detection. This study addresses this challenge by proposing an enhanced EfficientNet-YOLOv4 algorithm tailored explicitly for IC detection on PCBs. Integrating EfficientNet's robust feature extraction capabilities with YOLOv4's precise object localization, the methodology employs diverse data augmentation techniques to enrich the training dataset. Extensive experiments showcase the algorithm's efficacy and robustness in handling complex PCB layouts and varying lighting conditions, outperforming existing PCB inspection models. The proposed method, EfficientNetv2-L-YOLOv4, achieves an impressive F1-score of 99.22 with an inference speed of 0.135 seconds. The proposed method also performed well compared to EfficientNet-B7-FasterRCNN and original YOLOv4; it attains an F1-score of 98.96 and an inference speed of 0.102 seconds (with a batch size of 4). These results underscore the significance of effective feature extraction networks in object detection. Beyond addressing IC detection challenges, this algorithm advances computer vision and object detection fields. Implementing EfficientNetv2-L-YOLOv4 in real manufacturing scenarios holds promise for automating component inspection, potentially eliminating human intervention. INDEX TERMSAutomated visual inspection, Feature extraction networks, Object detection, Printed circuit board (PCB).

Implementation and Evaluation of Spatial Attention Mechanism in Apricot Disease Detection Using Adaptive Sampling Latent Variable Network

“…RetinaNet, utilizing a feature pyramid and focal loss function, performs well on datasets with class imbalances but may not reach optimal performance in tasks requiring high local detail recognition such as apricot tree disease detection, possibly due to limitations of its inherent structure and loss function. EfficientDet [ 61 ] displayed slightly superior performance to RetinaNet, with a precision of 0.84, a recall of 0.82, an accuracy of 0.83, and an mAP of 0.84. EfficientDet’s innovative use of BiFPN and compound scaling techniques optimizes the learning of multi-scale features, which is highly beneficial for handling various sizes of disease spots on apricot trees.…”