A space and time efficient convolutional neural network for age group estimation from facial images

Area of Science:

• Computer Vision
• Artificial Intelligence
• Biometrics

Background:

• Facial age estimation is crucial for applications like security, human-computer interaction, and biometrics.
• Facial aging is influenced by genetics, lifestyle, and environmental factors, making accurate prediction challenging.
• Deep learning, particularly Convolutional Neural Networks (CNNs), has shown promise in facial age estimation by automatically extracting features.

Purpose of the Study:

• To propose a space and time-efficient CNN method for extracting facial features and classifying age groups.
• To investigate CNN structures optimized for smaller image sizes and assess their impact on performance.
• To improve the accuracy and efficiency of facial age group classification.

Main Methods:

• Developed a space and time-efficient CNN architecture incorporating sufficient convolution layers to compensate for performance loss with fewer parameters.
• Designed and tested CNN structures capable of processing reduced image sizes to evaluate the effect of size reduction.
• Utilized deep learning for automatic extraction of distinct facial features for age classification.

Main Results:

• The proposed CNN method was validated on the UTKFace and Facial-age datasets.
• The model demonstrated superior classification accuracy compared to recent studies.
• Achieved an overall weighted F1-score of 87.84% for the age-group classification task.

Conclusions:

• The developed CNN method offers an efficient and accurate approach for facial age group classification.
• The study highlights the effectiveness of optimized CNN structures for handling reduced image sizes in facial analysis.
• The findings contribute to advancing the field of AI-driven biometrics and facial recognition systems.