Malnutrition Detection Using Deep Learning

Malnutrition in children remains a pressing global concern, affecting physical growth, cognitive development, and overall well-being. Early detection is critical for timely interventions that can help mitigate irreversible health consequences. This project proposes a computer vision approach leveraging Deep Learning techniques to distinguish between healthy children and those experiencing malnutrition from image data. By employing Convolutional Neural Networks (CNN), MobileNet, and VGG16 architectures, our system automatically classifies pediatric facial or bodily images into two categories—healthy versus malnourished. The dataset, sourced from Kaggle, comprises diverse images that capture various nutritional statuses across different regions and ethnicities. The primary objective is to accurately identify malnutrition signs in children based on physical features, enabling healthcare professionals and organizations to focus on those at the highest risk. This methodology integrates data preprocessing, feature extraction, model training, and validation steps to ensure robust predictive performance. The system’s effectiveness can potentially accelerate field screenings, empower telemedicine applications, and bolster policy decisions aimed at curtailing child malnutrition. Moreover, the research explores model optimization techniques and highlights the scalability of this classification approach for broader public health initiatives.

Keywords: Malnutrition Detection, Deep Learning, CNN, MobileNet, VGG16, Pediatric Health, Image Classification, Nutrition, Healthcare, Kaggle