TOWARDS SEED SELECTION AND YIELD ASSESSMENT FOR AGRICULTURAL PRODUCTIVITY IN INDIA

Abstract

In India, agriculture faces challenges such as climatic change and water scarcity, hindering farmers' ability to meet the high demand for products such as rice. To address this, a research project focused on seed selection and yield assessment, which is crucial, factors affecting production. Four rice varieties commonly cultivated in Tamil Nadu were chosen for experimentation: KO50, Atchaya Ponni, Andhra Ponni, and IR 20. The proposed method employs a machine vision system to measure seed quality and detect adulteration rates using various deep learning techniques. Real-time datasets and economically feasible imaging devices were used in this study. This includes the application of various deep learning techniques, with InceptionV3 exhibiting the highest accuracy at 98.96%, followed by ResNet101 at 86.61%. Convolutional Neural Network (CNN), AlexNet, and MobileNet also demonstrated respectable accuracies of 85.12%, 83.83%, and 81.99%, respectively. This research aims to empower farmers with tools to select high-quality seeds, potentially improving crop yield, and addressing production challenges in agriculture.