Scholarlink

<p>Facial recognition systems are becoming increasingly important in everyday tasks. They serve as a crucial element in recognition research and are necessary for human–machine interactions. These systems are extremely sensitive, and it may be challenging to recognize their original face due to their shifting points of view. This work suggests a novel deep learning technique that appears to be a good strategy for facial recognition in an uncontrolled environment because of its high degree of accuracy. Initially, an Improved Kuan Filter method is used to pre-process the input images to remove unnecessary information and improve the model’s performance. Then, the Attention-Based Deep Convolutional Neural Network (ADCNN), which improves the learning process and allows for the creation of better models with less computation, is used to extract the features. The Improved Red Fox Optimization (IRFO) algorithm selects the most effective features. Finally, face recognition is classified&#xa0;by introducing a new SVM-based ConvFaceNeXt (SVM_CFN) method. As a result, the proposed study enhances efficiency in categorizing different face recognition classes in an unconstrained environment. For simulation, the proposed method prefers the Python programming language, and results are analyzed using CASIA-Webface and LFW datasets. The simulation results show that the proposed model provides better results than other existing models in terms of accuracy of 99.7% for the CASIA-Webface and 99.6% for the LFW dataset.</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

An attention based deep learning with effective SVM-ConvFaceNeXt model for face recognition in unconstrained environment

  • 期刊论文

  • 发表日期：

    2025年8月22日

查看全文
  • Siriki Atchuta Bhavani,
  • C. Karthikeyan

摘要

Facial recognition systems are becoming increasingly important in everyday tasks. They serve as a crucial element in recognition research and are necessary for human–machine interactions. These systems are extremely sensitive, and it may be challenging to recognize their original face due to their shifting points of view. This work suggests a novel deep learning technique that appears to be a good strategy for facial recognition in an uncontrolled environment because of its high degree of accuracy. Initially, an Improved Kuan Filter method is used to pre-process the input images to remove unnecessary information and improve the model’s performance. Then, the Attention-Based Deep Convolutional Neural Network (ADCNN), which improves the learning process and allows for the creation of better models with less computation, is used to extract the features. The Improved Red Fox Optimization (IRFO) algorithm selects the most effective features. Finally, face recognition is classified by introducing a new SVM-based ConvFaceNeXt (SVM_CFN) method. As a result, the proposed study enhances efficiency in categorizing different face recognition classes in an unconstrained environment. For simulation, the proposed method prefers the Python programming language, and results are analyzed using CASIA-Webface and LFW datasets. The simulation results show that the proposed model provides better results than other existing models in terms of accuracy of 99.7% for the CASIA-Webface and 99.6% for the LFW dataset.