A Deep Learning Approach to Predict Chronological Age.

Area of Science:

• Computer Vision
• Biometrics
• Artificial Intelligence

Background:

• Accurate age prediction is crucial for various applications, including medical diagnostics (e.g., Alzheimer's disease) and public health policies (e.g., vaccination strategies).
• Existing age estimation methods often struggle with accuracy due to variations in facial features like shape, pose, and scale.
• Saudi Arabia's Vision 2030 emphasizes improving quality of life, with a focus on age-based health initiatives for the elderly.

Purpose of the Study:

• To propose a practical, consistent, and trustworthy method for real-time age prediction.
• To leverage the informative features of eye color intensity for enhanced age estimation.
• To develop an age prediction system that overcomes the limitations of current facial recognition approaches.

Main Methods:

• Utilized a segmentation algorithm to extract eye regions from images or video streams.
• Employed an ensemble of Convolutional Neural Networks (CNNs) trained on a large dataset (270,000+ images, ages 4-59).
• Focused on analyzing the color intensity of the eyes as a primary feature for age prediction.

Main Results:

• The proposed method achieved a high accuracy of 97.29%.
• The system demonstrated a Mean Square Error (MSE) of ±8.69 years.
• Comparative evaluation showed superior performance in accuracy, MSE, and Mean Absolute Error (MAE) against relevant studies.

Conclusions:

• Eye color intensity is a highly effective feature for accurate age prediction.
• The developed CNN-based ensemble method offers a reliable and accurate solution for real-time age estimation.
• The approach shows significant potential for real-world applications, particularly in healthcare and age-based policy implementation.