Lists
Lists
Bahrami, Samaneh, Hassan, Bilal, Das, Sonjoy Ranjon and Patel, Preeti (2025) Transfer learning based age recognition using arbitrary images. In: 4th 2025 IEEE World Conference on Applied Intelligence and Computing (AIC 2025), 26-27 July 2025, The Institution of Electronics and Telecommunication Engineers (IETE), Delhi Centre, India. (In Press)
Age recognition is an important task in computer vision and is applied in security, human computer interaction, and demographic analysis. The purpose of this study is to examine the benefits of transfer learning that enables age recognition models to benefit from arbitrary image data. To test the impact of (CNNs) have on increasing the accuracy of age estimation for various image datasets, they were evaluated. Through the analysis of widely used deep learning architectures, VGG16, VGG19, ResNet50, and MobileNet, this study aims to select the most preliminary and optimization techniques for age classification. Fine-tuned ImageNet pretrained models were applied in transfer learning to adapt these models for age recognition tasks. Supplementing the UTKFace dataset, we investigated the influence of different CNN architectures on model performance, adding to the data carefully curated primary dataset. In this study, it was determined that VGG16 achieved an optimal balance between computational efficiency and accuracy, yielding results of 80% on the primary dataset and 71% on the UTKFace dataset when compared with all four models. This study explores the effectiveness of deep learning models for age classification, focusing on the VGG16 architecture. By leveraging transfer learning techniques and diverse datasets, the model achieves robust performance across varying demographics. The results highlight the significance of dataset diversity and advanced learning strategies in developing scalable and adaptable age recognition systems for real-world deployment.
Available under License Creative Commons Attribution 4.0.
Download (1MB) | Preview
 |
View Item |