Scholarlink

<p>A major challenge in facial recognition is maintaining high accuracy without requiring individuals to remove their face masks. This research focuses on recognizing faces by analyzing the upper half—specifically the eyes, forehead, and visible side features like ears. We evaluated nine masked face recognition techniques, including MobileNet, MobileNetV2, ResNet50, DenseNet121, EfficientNetB0, InceptionV3, ResNet101V2, Xception, and our proposed model, using a dataset of ten classes. Our methodology introduces a novel approach that combines a modified MobileNetV2 with a convolutional architecture for fast feature embedding (CAFFE). This design was developed after extensive study of face recognition methods for artificial intelligence applications. To support this, we created a dataset of 3,340 masked face images using artificial masking and data augmentation techniques. The dataset was used to train and test our Python-based AI model. The core objective was to improve the model’s performance by enhancing a pre-trained structure with 7 additional layers, optimized weights, and associated configuration files. Our proposed model integrates transfer learning, deep neural network (DNN) and deep learning (DL) architecture to extract image features of masked face classification (MFC) and masked face recognition (MFR) strategies to effectively extract features for masked face classification and recognition. In conclusion, our model outperformed the other eight traditional approaches, achieving 98.34% accuracy with a 1.66% error rate, while offering fast computation and reduced parameter usage. The source code used in our article are now made available on github platform <a href="https://github.com/AnilKumargithu/Masked-Face-Recognition-MFR-Classification-and-Prediction">https://github.com/AnilKumargithu/Masked-Face-Recognition-MFR-Classification-and-Prediction</a>. Our created dataset of artificially masked face images utilized in this research is made available on Kaggle <a href="https://www.kaggle.com/datasets/banilkumar20phd7071/masked-face-recognition-dataset-artificial-masking">https://www.kaggle.com/datasets/banilkumar20phd7071/masked-face-recognition-dataset-artificial-masking</a>. </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

A deep learning-based novel model for masked face recognition

  • 期刊论文

  • 发表日期：

    2025年9月15日

查看全文
  • B. Anil Kumar,
  • Neeraj Kumar Misra

摘要

A major challenge in facial recognition is maintaining high accuracy without requiring individuals to remove their face masks. This research focuses on recognizing faces by analyzing the upper half—specifically the eyes, forehead, and visible side features like ears. We evaluated nine masked face recognition techniques, including MobileNet, MobileNetV2, ResNet50, DenseNet121, EfficientNetB0, InceptionV3, ResNet101V2, Xception, and our proposed model, using a dataset of ten classes. Our methodology introduces a novel approach that combines a modified MobileNetV2 with a convolutional architecture for fast feature embedding (CAFFE). This design was developed after extensive study of face recognition methods for artificial intelligence applications. To support this, we created a dataset of 3,340 masked face images using artificial masking and data augmentation techniques. The dataset was used to train and test our Python-based AI model. The core objective was to improve the model’s performance by enhancing a pre-trained structure with 7 additional layers, optimized weights, and associated configuration files. Our proposed model integrates transfer learning, deep neural network (DNN) and deep learning (DL) architecture to extract image features of masked face classification (MFC) and masked face recognition (MFR) strategies to effectively extract features for masked face classification and recognition. In conclusion, our model outperformed the other eight traditional approaches, achieving 98.34% accuracy with a 1.66% error rate, while offering fast computation and reduced parameter usage. The source code used in our article are now made available on github platform https://github.com/AnilKumargithu/Masked-Face-Recognition-MFR-Classification-and-Prediction. Our created dataset of artificially masked face images utilized in this research is made available on Kaggle https://www.kaggle.com/datasets/banilkumar20phd7071/masked-face-recognition-dataset-artificial-masking.