Agricultural Pest and Disease Detection System using Machine Learning

Crop diseases pose a significant threat to agricultural productivity in Manipur, a state renowned for its rich agricultural heritage and diverse crop cultivation practices. Early detection of these diseases is crucial for safeguarding the region's agricultural sustainability and ensuring food security. In this paper, we present a comprehensive machine learning-based system utilizing Convolutional Neural Network (CNN) and the U-NET model for detecting crop diseases specifically tailored to the unique agricultural landscape of Manipur, India. Leveraging advanced techniques such as transfer learning with deep learning pre-trained models, including VGG16 and VGG19, we developed a robust system capable of accurately identifying diseased crops, including locally grown vegetables like "Kobiphun" (cabbage), "Kobilei" (cauliflower), "Aloo" (potato), "Khamen Asinba" (tomato), and "Chak-hao" (black rice). The datasets used in this project are purely self-collected primary data, reflecting extensive effort and dedication in gathering high-quality images for training and model development. Our methodology involves detailed image processing, feature extraction demonstrating the effectiveness of our approach in addressing the specific challenges posed by crop diseases in Manipur's agricultural system. The findings highlight the model's high accuracy (over 92%), precision (exceeding 90%), and recall (approaching 100%), indicating its reliability in real-world applications. Additionally, the system's scalability allows for integration with real-time monitoring tools, enhancing its applicability for local farmers. This research contributes to improved disease management practices, enhances the resilience of Manipur's agriculture sector, and ensures the livelihoods of its farming communities.