AI-Based Crop Disease Detection using Convolutional Neural Networks (CNN)

Abstract

Agricultural productivity faces significant threats from diseases that affect plant health and reduce crop yields substantially across global farming systems. The conventional approach to identifying crop diseases relies heavily on manual inspection by trained agricultural professionals, a process that is inherently time-consuming, lacks consistency, and cannot scale to meet the requirements of modern large-scale farming operations. This study introduces and evaluates an artificial intelligence framework leveraging Convolutional Neural Networks to automate the detection and classification of crop diseases from digital images of affected plant leaves. Employing the extensively used PlantVillage benchmark dataset encompassing over 50,000 annotated images across multiple crop varieties and corresponding disease categories, the developed system demonstrates classification performance of approximately 88-90% accuracy with inference time below three seconds per image. The approach integrates image acquisition, sophisticated preprocessing techniques, deep neural network classification, and automated recommendation generation to support farming communities. The system design prioritizes accessibility for resource-limited farmers through simplified interfaces and mobile-first deployment strategies. Results indicate that machine learning-based automated detection offers significant potential for advancing agricultural sustainability, reducing reliance on broad-spectrum chemical treatments, and supporting informed decision-making by farming communities worldwide.