GENDER DETECTION FROM A FACE MASK BY USING DEEP LEARNING ALGORITHM

Abstract

Accurately identifying gender from masked facial images is a critical challenge, especially during widespread mask usage in pandemics. This study introduces a novel deep learning approach that combines convolutional architectures with transfer learning to achieve robust gender classification in masked faces. Using a curated dataset of 89,131 masked facial images for training and 20,714 for evaluation, the proposed method outperformed 11 traditional models, achieving 99% accuracy and a 0.17% error rate with minimal computational resources. By leveraging transfer learning, the framework achieves a balance between efficiency and accuracy, making it suitable for real-time applications. Comparative evaluations demonstrate the superiority of the proposed model over traditional techniques like MobileNetV2, ResNet101, and DenseNet169. This research provides insights into enhancing the accuracy of facial recognition systems and improving biometric classification in constrained environments.