Advanced CNN-Based Leaf Disease Detection for Sustainable Agriculture

Abstract

Crop diseases are one of the major challenges to global agricultural productivity, and efficient and accurate solutions for early detection are needed. This paper proposes a CNN-based model for crop leaf disease detection, integrated with the Crop Pro user interface for practical deployment. The model is trained on an extensive dataset that contains 8,264 images of crop leaves belonging to various species, classified into 14 disease categories, with preprocessing and augmentation techniques further improving it. The CNN model had a training accuracy of 94.73% and a validation accuracy of 88.75%, showing robustness and the ability to generalize over different crops and conditions. Users can upload their leaf images to the Crop Pro platform that preprocesses and classifies images and uses them for high-confidence disease predictions. The availability of an iterative feedback loop helps improve the model based on input received from the user. Scalability of the system and application in real-world applications may offer promising solutions for precision agriculture through early intervention for sustainable crop management.