Age group classification with OpenCV and deep learning on Raspberry Pi

This paper presents a lightweight and efficient real-time system for age group and gender classification using deep learning techniques on an embedded edge device Raspberry Pi 4. The proposed system captures live facial images through a camera, detects faces, preprocesses them, and classifies the age and gender using one of three deep learning models: a fast Caffe model, an accurate DeepFace model, and a custom-trained Convolutional Neural Network (CNN) model optimized for Raspberry Pi. The system operates entirely offline, making it suitable for privacy-sensitive environments such as retail, education, and security sectors. Experimental results show that the custom CNN model achieves superior performance with 83.7% gender classification accuracy and an age prediction mean absolute error (MAE) of 5.1 years. The results validate the feasibility of deploying deep learning models for demographic analysis in low-cost, real-time, and portable edge computing environments. The study also compares the trade-offs between speed and accuracy among the three models, highlighting the custom CNN's suitability for real-world applications without reliance on cloud infrastructure.