Addressing Data Imbalance in Plant Disease Recognition through Contrastive Learning

Abstract

The following study introduces a novel framework for recognizing plant diseases, tackling the issue of imbalanced datasets, which is prevalent in agriculture, a key sector for many economies. Plant diseases can significantly affect crop quality and yield, making early and accurate detection vital for effective disease management. Traditional Convolutional Neural Networks (CNNs) have shown promise in plant disease recognition but often fall short with non-tomato crops due to class imbalance in datasets. The proposed approach utilizes contrastive learning to train a model on the PlantDoc dataset in a self-supervised manner, allowing it to learn meaningful representations from unlabeled data by maximizing the similarity between images based on disease state rather than species. This method shows a marked improvement in accuracy, achieving 87.42% on the PlantDoc dataset and demonstrating its superiority over existing supervised learning methods. The agnostic nature of the model towards plant species allows for universal application in agriculture, offering a significant tool for disease management and enhancing productivity in both existing farms and future smart farming environments.