Scholarlink

Facial recognition is a biometric identification technology that analyzes and compares the unique facial features of individuals in images or video frames to verify their identity. Such systems utilize advanced pattern recognition and machine learning techniques to examine characteristics like the distance between the eyes, the shape of the nose, and the contour of the jawline, matching them against a database of known faces. This study evaluates the performance of facial recognition models under challenging conditions, including low illumination, partial occlusions, and variations in facial expressions. It also explores the integration of 3D face reconstruction and multimodal biometric technologies to enhance identification accuracy and scalability. The research focuses on analyzing the architecture and performance of advanced variants of convolutional neural networks (CNN), including DenseNet169, DenseNet121, and EfficientNetV2B0.This paper introduces a facial recognition technology that is based on deep learning. The experiment’s highest accuracy was 99.52% for DenseNet169, 99.13% for DenseNet121, and 99.33% for EfficientNetV2B0. The significance of model combination in enhancing the stability of recognition in various real-world environments is underscored by the maximum accuracy of 99.86% obtained by an ensemble method that combines these three models.

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

Enhancing Face Recognition in Unconstrained Conditions Using Ensemble Deep Learning Models

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • D. M. Asadujjaman,
  • Md. Mahfujur Rahman,
  • Md. Mahmuduzzaman Ikbal,
  • Abdullah Tamim,
  • Mst. Jannatul Ferdous

摘要

Facial recognition is a biometric identification technology that analyzes and compares the unique facial features of individuals in images or video frames to verify their identity. Such systems utilize advanced pattern recognition and machine learning techniques to examine characteristics like the distance between the eyes, the shape of the nose, and the contour of the jawline, matching them against a database of known faces. This study evaluates the performance of facial recognition models under challenging conditions, including low illumination, partial occlusions, and variations in facial expressions. It also explores the integration of 3D face reconstruction and multimodal biometric technologies to enhance identification accuracy and scalability. The research focuses on analyzing the architecture and performance of advanced variants of convolutional neural networks (CNN), including DenseNet169, DenseNet121, and EfficientNetV2B0.This paper introduces a facial recognition technology that is based on deep learning. The experiment’s highest accuracy was 99.52% for DenseNet169, 99.13% for DenseNet121, and 99.33% for EfficientNetV2B0. The significance of model combination in enhancing the stability of recognition in various real-world environments is underscored by the maximum accuracy of 99.86% obtained by an ensemble method that combines these three models.