Deep Learning Techniques for Plant and Soybean Leaf Disease Detection: A Review of Models, Datasets, Gaps, Challenges and Future Directions

Abstract

Plant diseases have a high impact on agricultural productivity and food security in the world. Plant diagnosis is also critical in avoiding a loss of crops and enhancing agricultural productivity by timely detecting plant diseases. The conventional methods of identifying diseases are by manual inspection and expert knowledge, which are time consuming and have the risk of human error. Over the last few years, deep learning and computer vision methods have evolved as powerful solutions towards automated detection of plant diseases using leaf images. The demand for soybeans as a source of oil and protein on the world market makes them one of the most important crops in the world. Because of this, researchers are now increasingly interested in developing deep learning systems to detect soybean leaf diseases. This paper reviews the recent advancements in the application of deep learning techniques for the identification of plant and soybean leaf diseases. It presents multiple deep learning structures such as convolutional neural networks (CNNs), transformer-based neural networks, hybrids of deep learning, and ensemble techniques. Additionally, the paper discusses the importance of high-quality training data and surveys publicly available datasets used in soybean disease detection. Other important issues analyzed in the study include constraints of the dataset, environmental changes, complexity in computing, and interpretability of the model. Lastly, the paper describes the possible future research directions to enhance the accuracy, robustness, and scalability of automated plant disease detection systems. This review is an overview of the existing studies in the field of deep learning-based detection of plant diseases and an introduction to the study by the prospective researcher in the area of smart agriculture and precision farming technologies.