An Automated Deep Learning Framework for Millet Classification Using Convolutional Neural Networks

Abstract

Despite the nutritional benefits of millets, many people lack awareness of their importance in a balanced diet. This paper proposes a millet recognition system for image classification with the use of various deep learning techniques. In this paper, a system has been built with deep learning models using Convolutional Neural Network (CNN) to automatically classify the various millet types. Also, it provides comprehensive information, benefits, nutrition charts, and recipes. A custom-made dataset of high-quality images of the various types of millets is used. Various people from different regions and age groups have also been surveyed to learn about their knowledge of millets. The aim is to educate users about the nutritional values and health benefits of millets. Experimented with the models such as Residual Neural Network (ResNet-50), Visual Geometry Group (VGG-16), ALexNet, MobileNetV2, the best performing model out of all was the VGG-16 which gave an accuracy of 99.29%.