Deep CNN Model: A case study of predicting security surveillance activities utilizing Gender and Age

Abstract

In this paper, we present F-AgeNet, a novel and highly efficient convolutional neural network (CNN) model tailored for the task of age and face detection. Leveraging a combined dataset comprising 22,708 images from a public dataset and 110 raw images from a private dataset, FAgeNet demonstrates remarkable accuracy in age group classification. The proposed model outperforms widely recognized models such as VGGFace, OpenFace, DeepFace, EfficientNet, and MobileNetV2, achieving a final test accuracy of 88.97%. Our age classification system categorizes individuals into four distinct groups: Group 1 for ages between 0 and 18, Group 2 for ages under 30, Group 3 for ages under 80, and Group 4 for individuals aged 80 and above. This granular age grouping not only enhances the model's precision but also provides valuable insights into agerelated facial features. F-AgeNet's architecture is meticulously designed to address the challenges associated with both face and age detection. Through a careful fusion of the public and private datasets, our model gains a comprehensive understanding of diverse facial characteristics, contributing to its robust performance. The utilization of 110 raw images from the private dataset further enriches the training process, making F-AgeNet adept at handling real-world scenarios. Comparative analysis with existing state-of-the-art models reveals the superiority of F-AgeNet in achieving high accuracy. The model's success can be attributed to its ability to extract intricate facial features and discern subtle age-related patterns. The experimental results showcase FAgeNet's capability to surpass benchmark models, making it a valuable addition to the domain of age and face detection. In addition to presenting F-AgeNet's superior performance, we contribute a comprehensive evaluation of various established models, including VGGFace, OpenFace, DeepFace, EfficientNet, and MobileNetV2. Our findings not only highlight F-AgeNet's efficacy but also provide insights into the strengths and limitations of existing models in the context of age and face detection.