Age and Gender Identification Using Neural Image Processing

Authors- Tanmay Hadke

Abstract-The rapid advancement in the fields of computer vision and deep learning has enabled out- standing achievements in the recognition of facial attributes such as age and gender. This report is about developing a useful system that uses neural image processing methods for real-time age and gender classification. The proposed system uses two distinct methods: one by using OpenCV’s ‘resize‘ function to resize the image and the other using Pillow library’s ‘Image.resize‘ method. These preprocessing techniques are critical to preparing facial image data to feed into a chosen MobileNetV2-based neural network architecture, selected for its lightweight design and efficiency in computation. This research extensively evaluates the two different preprocessing methods to focus on their effect on accuracy, computation time, and resource usage. Optimized MobileNetV2 architecture is used to classify age and gender using facial datasets available in the public domain for training purposes. The models are then tested using the webcam for real-time input to analyze their usefulness in practice. This project presents a comparative analysis of two different feature extraction techniques to determine the best preprocessing method for neural network-driven age and gender detection systems. Relevant conclusions regarding the interplay between preprocessing methodologies and model efficacy can be obtained from the results and be put forward towards lightweight, accurate, and resource-conserving demographic analysis systems that can be success- fully applied in different real-life scenarios. Keywords: Age and Gender Classification, Neural Image Processing, MobileNetV2, OpenCV, Pillow Library, Feature Extraction, Deep Learning, Computer Vision.

DOI: /10.61463/ijset.vol.13.issue1.128