Detection of Malnutrition in Children Under 5 Years of Old Using Deep Learning

Abstract

This study evaluates the performance of Convolutional Neural Networks (CNNs) for the early detection of childhood malnutrition. The architectures analyzed include ResNet-50, EfficientNet-B4, VGG16, and AlexNet, selected for their demonstrated effectiveness in image classification tasks. Images were categorized into three classes - normal, at risk of malnutrition, and severely malnourished - and underwent preprocessing steps such as resizing, flipping, zooming, and face detection using MTCNN. The evaluation employed metrics including accuracy, precision, recall, and F1-score, with ResNet-50 emerging as the most effective model, achieving an accuracy of 92%. Based on these findings, this study explores the potential application of ResNet-50 in mobile solutions to provide accessible and practical tools for malnutrition detection in resource-limited settings.