Scholarlink

In recent decades face recognition has emerged as an important research area in computer vision and remains one of the most successful applications of image analysis and understanding This problem a complexity has attracted the interest not only of computer scientists but also of neuroscientists and psychologists. Many believe that advances in computer vision can provide valuable insights into how the human brain works, and vice versa. One of the most critical duties in the field of computer vision is the identification of individuals through facial recognition. Although numerous studies have concentrated on balanced, enhanced, or enhanced data, only a small number have examined genuine “wild-life” images. In this paper, we propose a deep learning algorithm that uses a CNN-based feature extraction method combined with principal component analysis (PCA), We evaluate CNN classifiers on the Labeled Faces in the Wild (LFW) dataset including K-Nearest Neighbors (K-NN), Naïve Bayes (NB), and AdaBoosted Support Vector Machines (SVM). This process, which involved gathering, pre-processing and managing a great number of face images ensured the preservation of face images of important and diverse species, while lowering the dataset spatially. Using a multi-step approach pre-processing produced a high quality, noninferior face image dataset. The resulting low-density facial image dataset, enriched by dimension reduction techniques such as PCA, effectively reduces the size of CNN feature maps while preserving informative features This dataset provides researchers and professionals gain a valuable resource for developing robust and flexible computer vision applications. The CNN model was trained using a variety of classifiers that were evaluated on a data set of 13,233 images using fivefold cross-validation (CV). The findings indicate that PCA is a viable approach for the reduction of dimensionality and the extraction of features in face recognition. Among the supervised learning methods used, SVM, when properly configured, obtained the highest accuracy for this problem.

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

Optimizing Deep Leaning Approach for a Significant Low-Dimensional Face Image Recognition Using Boosted and Non-boosted Classifiers

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Adil Al-Azzawi

摘要

In recent decades face recognition has emerged as an important research area in computer vision and remains one of the most successful applications of image analysis and understanding This problem a complexity has attracted the interest not only of computer scientists but also of neuroscientists and psychologists. Many believe that advances in computer vision can provide valuable insights into how the human brain works, and vice versa. One of the most critical duties in the field of computer vision is the identification of individuals through facial recognition. Although numerous studies have concentrated on balanced, enhanced, or enhanced data, only a small number have examined genuine “wild-life” images. In this paper, we propose a deep learning algorithm that uses a CNN-based feature extraction method combined with principal component analysis (PCA), We evaluate CNN classifiers on the Labeled Faces in the Wild (LFW) dataset including K-Nearest Neighbors (K-NN), Naïve Bayes (NB), and AdaBoosted Support Vector Machines (SVM). This process, which involved gathering, pre-processing and managing a great number of face images ensured the preservation of face images of important and diverse species, while lowering the dataset spatially. Using a multi-step approach pre-processing produced a high quality, noninferior face image dataset. The resulting low-density facial image dataset, enriched by dimension reduction techniques such as PCA, effectively reduces the size of CNN feature maps while preserving informative features This dataset provides researchers and professionals gain a valuable resource for developing robust and flexible computer vision applications. The CNN model was trained using a variety of classifiers that were evaluated on a data set of 13,233 images using fivefold cross-validation (CV). The findings indicate that PCA is a viable approach for the reduction of dimensionality and the extraction of features in face recognition. Among the supervised learning methods used, SVM, when properly configured, obtained the highest accuracy for this problem.