Enhancing Rice Pest Classification Through Machine Learning and Deep Learning Techniques

Abstract

The production of rice is essential to the world's food security, yet pests provide a constant danger and can result in large productivity losses. For efficient pest management techniques, rice pests must be identified with precision and promptness. But farmers sometimes have trouble identifying crop illnesses early on, especially in rural regions where access to professional knowledge is scarce .In this paper, we offer a deep learning and machine learning (ML) based comprehensive strategy to rice pest categorization . The efficacy of five different models—ResNet150, CNN, VGG16 and VGG19,in categorizing photos of typical rice pests is assessed. The findings of the experiment show differences in the models' pest categorization accuracy. Although several models show encouraging results, others have drawbacks. In particular, CNN and VGG19 perform moderately with accuracy rates of 74.75 and 98.41%, respectively. ResNet50, on the other hand, has an impressive accuracy rate of 86.90%, demonstrating its potential for precise pest detection. VGG16 stands out as the best-performing model with an astounding accuracy rating of 98.47%. With an accuracy rate of 98.47%, VGG16 closely trails, proving their dominance in jobs involving the categorization of rice pests. These results emphasize how crucial it is to choose the right deep learning architectures for precise and effective insect identification in rice farming. The excellent accuracy rates attained by VGG19 and VGG16 indicate that they are suitable for real-world use in agricultural environments. Through the facilitation of prompt identification and focused control of rice pests, our machine learning framework can enable farmers to reduce crop losses and advance sustainable rice production.