Scholarlink

<p>Examination malpractice has eroded moral integrity and ethical principles within many Nigerian institutions. Traditional identification methods, like student ID cards and fee payment receipts, are easily falsifiable, fostering impersonation during examination. This research proposes a convolutional neural network (CNN)-based facial recognition system to address this issue. A facial corpus of students was collected, stored in the database. Uploaded media was saved in Amazon Web Services (AWS) and processed using Open Computer Vision (OpenCV) library. Feature extractor and facial recognition are based on MobileNetV2, ResNet50 and EfficientNet-B0, which aid the translation of incoming images. The database (MySQL) facilitates the retrieval of the data during the verification process. The system was implemented using JavaScript and the Python programming platform (Python Web Framework). In our experiments, MobileNetV2 achieved an overall accuracy of 96%, with precision of 99%, recall of 92%, and an F1-score of 95%. ResNet50 accuracy measure 91%, precision of 93%, Recall of 88% and F1-score of 90% while EfficientNet-B0 accuracy measure 92%, precision of 1%, recall of 83% and F1-score of 91%. These results demonstrate that MobileNetV2 provides high reliability, making it ideal for deployment on resource-constrained devices. Hence, the proposed system is recommended for educational institutions to effectively curb impersonation and enhance examination integrity.</p>

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

Convolution Neural Network-Based Facial Recognition System for Detecting Student Impersonation in Educational Institution

  • 期刊论文

  • 发表日期：

    2026年2月12日

查看全文
  • Ednah Olubunmi Aliyu,
  • Adedamola Ayobami Adeleke,
  • Felix Ola Aranuwa,
  • Olanike Christianah Akinduyite,
  • Olaniyi Abiodun Ayeni

摘要

Examination malpractice has eroded moral integrity and ethical principles within many Nigerian institutions. Traditional identification methods, like student ID cards and fee payment receipts, are easily falsifiable, fostering impersonation during examination. This research proposes a convolutional neural network (CNN)-based facial recognition system to address this issue. A facial corpus of students was collected, stored in the database. Uploaded media was saved in Amazon Web Services (AWS) and processed using Open Computer Vision (OpenCV) library. Feature extractor and facial recognition are based on MobileNetV2, ResNet50 and EfficientNet-B0, which aid the translation of incoming images. The database (MySQL) facilitates the retrieval of the data during the verification process. The system was implemented using JavaScript and the Python programming platform (Python Web Framework). In our experiments, MobileNetV2 achieved an overall accuracy of 96%, with precision of 99%, recall of 92%, and an F1-score of 95%. ResNet50 accuracy measure 91%, precision of 93%, Recall of 88% and F1-score of 90% while EfficientNet-B0 accuracy measure 92%, precision of 1%, recall of 83% and F1-score of 91%. These results demonstrate that MobileNetV2 provides high reliability, making it ideal for deployment on resource-constrained devices. Hence, the proposed system is recommended for educational institutions to effectively curb impersonation and enhance examination integrity.