Scholarlink

The system proposed in the paper utilizes state-of-the-art deep learning methods, including ResNet-50, a deep convolutional neural network architecture. Faster R-CNN, an advanced object detection algorithm, is employed to efficiently divide the task of object detection into region proposal and classification stages. ResNet-50, known for its effectiveness in computer vision tasks, serves as the support network within the Faster R-CNN framework. Its deep layer structure, featuring residual blocks, addresses the vanishing gradient problem, enabling the extraction of intricate features from input images. This integrated approach allows for accurate detection and classification of defects on printed circuit boards (PCBs). The system’s accuracy, quantified through rigorous evaluation metrics, demonstrates its high precision, and recall rates in defect detection tasks, with an impressive accuracy rate of 95%. Renowned for its robustness and speed, the proposed method has gained widespread acceptance in research and industry, proving valuable for applications such as manufacturing quality control, surveillance anomaly detection, and medical image analysis.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Circuit Defect Inspect: Circuit Board Defect Detection Using Modified Faster R-CNN

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Jyoti Madake,
  • Nagnath Kotwal,
  • Aditya Kshirsagar,
  • Kaushal Kshirsagar,
  • Shripad Bhatlawande

摘要

The system proposed in the paper utilizes state-of-the-art deep learning methods, including ResNet-50, a deep convolutional neural network architecture. Faster R-CNN, an advanced object detection algorithm, is employed to efficiently divide the task of object detection into region proposal and classification stages. ResNet-50, known for its effectiveness in computer vision tasks, serves as the support network within the Faster R-CNN framework. Its deep layer structure, featuring residual blocks, addresses the vanishing gradient problem, enabling the extraction of intricate features from input images. This integrated approach allows for accurate detection and classification of defects on printed circuit boards (PCBs). The system’s accuracy, quantified through rigorous evaluation metrics, demonstrates its high precision, and recall rates in defect detection tasks, with an impressive accuracy rate of 95%. Renowned for its robustness and speed, the proposed method has gained widespread acceptance in research and industry, proving valuable for applications such as manufacturing quality control, surveillance anomaly detection, and medical image analysis.