Age Stage Prediction Using Light Convolution Neural Network

Abstract

Age-stage prediction models based on facial images are essential for various applications, including forensic investigations, medical imaging, social media advertising, and compliance with age-related regulations. These models provide valuable insights and predictions, helping inform decision-making and improve outcomes. The proposed model estimates the human age stage by analyzing a face image, extracting various facial features and wrinkles, and then connecting these extracted feature patterns to determine the age stage. To address potential limitations in facial images, such as variations in lighting, pose, expression, and image quality, which can impact prediction performance, the proposed model utilizes a neural network architecture. On the other hand, the proposed CNN model has adopted a new light architecture to make it usable with reasonable computational capabilities. The practical results of this model resulted in an accuracy of 98.86% when implemented on the UTKFace dataset, which consists of over 20,000 face images.