Balanced-YOLOv3: Addressing the Imbalance Problem of Object Detection in PCB Assembly Scene

Abstract

The object detection algorithm of the PCB (Printed Circuit Board) assembly scene based on CNN (Convolutional Neural Network) can significantly improve the production capacity of intelligent manufacturing of electronic products. However, the object class imbalance in the PCB assembly scene, the multi-scale feature imbalance, and the positive/negative sample imbalance in the CNN have become critical problems restricting object detection performance. Based on YOLOv3, this paper proposes a class-balanced Train/Val (Training set/Validation set) split method for object class imbalance, an additional feature fusion strategy for multi-scale feature imbalance, and an efficient anchor concept for positive/negative sample imbalance. These three contributions are Balanced-YOLOv3. After experimental verification, compared with other YOLOv3 series algorithms, the mAP@.5 (Mean Average Precision at Intersection over Union threshold 0.5) and mAP@.5:.95 (average mAP over different Intersection over Union thresholds, from 0.5 to 0.95, step 0.05) of Balanced-YOLOv3 have achieved the best results and ranked third in the metrics of parameter and inference time. Compared with other current anchor-based object detection algorithms, Balanced-YOLOv3 has excellent detection performance and low computational complexity, which effectively solves the problem of imbalanced object detection in PCB assembly scenarios.

Improved YOLOv3 Model for Workpiece Stud Leakage Detection

“…Fi,j = W BN × F i,j + b BN (10) In Equation ( 10), Fi,j denotes the normalized result; W BN represents the BN layer weight; b BN represents the BN layer bias; and F i,j represents the feature map processed by the convolutional layer, which is represented as follows.…”

“…This advantage effectively avoids the problem of traditional recognition methods that require expert knowledge and manual configurations. In recent years, convolutional neural networks (CNNs) [9] have been extensively used in the field of workpiece stud image recognition and target detection [10,11]. Liu et al [12] proposed an online stud measurement method based on photometric stereo measurements and deep learning theory.…”

PCBDet: An efficient deep neural network object detection architecture for automatic PCB component detection on the edge

“…The ability to automatically detect different electronic components on a PCB for automated inspection purposes is thus of great value. As a result, there has been significant interest in the research community in automatic PCB component detection, which involves the identification of the type and location of a PCB's components, as pictured in Figure 1, particularly leveraging deep learning [2,3]. One facet of this area of research that has largely been left unexplored, however, is that of computational efficiency, which is particularly critical for visual quality inspection, a task that typically operates in high-volume, high-throughput environments with limited edge computing resources.…”

PCBDet: An Efficient Deep Neural Network Object Detection Architecture for Automatic PCB Component Detection on the Edge

“…Therefore, the ability to automatically detect different electronic components on a PCB board for automated inspection purposes is highly desired. As a result, there has been significant interest in the research community in automatic PCB component detection, particularly leveraging deep learning [2,3]. However, one consideration that has been largely left unexplored in research literature in this area is computational efficiency, which is particularly critical for real-world visual quality inspection scenarios involving high-volume, high-throughput electronics manufacturing use-cases under constrained edge computing resources.…”