Multi-model learning system using EfficientNet classifier for malnutrition management

Many developed and developing countries still struggle with malnutrition, making it a global issue. Traditional methods to detect malnutrition are manual, slow, error-prone, and resource-intensive. This paper proposes a smart AI-based system to solve these problems. The system uses a phone camera to take a person&s;s photo, and detects key face and body points from the image. It uses two modules, EfficientNet for 2D image features, and PointNet for 3D body shape. The results from both are sent to another module (MLP), which checks the person&s;s health. The system then shows if the person is healthy or malnourished, and gives a diet plan if needed. It was tested with 300 to 2000 images. Even with 300 images, it gave 94% accurate results, and with more images, accuracy increased to 97.4%. This shows the system works well, quick, and can help people in both cities and rural areas.