Development of a Deep Learning Model for Age Estimation Utilizing Multimodal Feature and Augmented Analysis

Abstract

Age estimation from facial images is crucial in security, healthcare, and entertainment for applications like age-restricted content filtering and targeted advertising. However, this task is challenging due to pose, illumination, and aging variations. Traditional methods using handcrafted features often lack discriminative power. Deep learning, with its ability to learn complex features, has sparked interest in age estimation. This paper presents a deep learning-based approach for age estimation from frontal face images. It focuses on facial components like eyes, nose, and mouth to capture age-related changes. Augmentation techniques enhance robustness against variations, and a Binary Cascaded CNN with binary weights and activations reduces model complexity. The model transforms augmented components into multimodal features, allowing it to discern age-related changes across domains. Evaluated on FGNET and IMDB-WIKI datasets, the model achieves 99.5% accuracy with an MAE of 1.26 across all age groups. This high accuracy demonstrates the model's effectiveness and potential for real-world age estimation scenarios in diverse applications.