Agricultural Disease Prediction Using Image Analysis

Abstract

Plant diseases can cause significant damage to crops, impacting global food security and causing economic losses for farmers. Traditional methods of disease detection are time-consuming, expensive, and prone to errors. Deep learning models offer advantages by quickly and accurately analyzing large amounts of data, detecting patterns invisible to human experts. A study utilized the plant village dataset with over 50,000 images from 14 plant species. To ensure robustness, image augmentation techniques were applied. Models were trained on 8 plant species, including Apple, Tomato, Grape, Potato, Strawberry, Corn, Cherry, and Pepper, using 23,074 images from the dataset. The study classified 16 diseases and healthy leaves for the eight species. Performance of models trained with and without data augmentation was compared, considering accuracy, robustness, generalization capability, training time, and memory consumption. Additionally, a user-friendly interface was developed to make the system accessible to non-experts, enabling them to detect diseases in their plants without deep learning expertise. These advancements in efficient and accurate plant disease detection contribute to addressing food security concerns and supporting farmers in making informed decisions.