Comparative Analysis of VGG16 and EfficientNetB3 for Age Group Classification: A Study on Transfer Learning with Limited Data

Accurate age group classification from facial images has significant applications in various fields, including security, marketing, and healthcare. This research investigates the performance of two state-of-the-art deep learning architectures, VGG16 and EfficientNetB3, for age group classification. The study utilizes a limited dataset of facial images categorized into five age groups. Both models were fine-tuned using transfer learning to adapt to the specific task of age group classification. While VGG16 achieved a higher training accuracy, it suffered from overfitting, resulting in poor validation and test performance. In contrast, EfficientNetB3, despite a lower training accuracy, demonstrated superior generalization ability, achieving a higher test accuracy. The findings highlight the challenges associated with training deep learning models on limited datasets. To improve the performance of age group classification models, future research may explore techniques such as data augmentation, regularization, and more advanced model architectures.