Systematic study on deep learning-based plant disease detection or classification

• Published in: The Artificial intelligence review Vol. 56; no. 12; pp. 14955 - 15052
• Main Authors: Sunil, C. K., Jaidhar, C. D., Patil, Nagamma
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.12.2023; Springer; Springer Nature B.V
• Subjects: Agricultural production; Analysis; Artificial Intelligence; Classification; Computational linguistics; Computer Science; Datasets; Deep learning; Economic impact; Grain; Language processing; Machine learning; Medical research; Medicine, Experimental; Natural language interfaces; Operating costs; Plant diseases; Subject specialists; Transportation; Travel time; Smart agriculture; Climate change; Plant pathology; Deep Learning model; Deep Convolutional Neural Network
• ISSN: 0269-2821; 1573-7462
• DOI: 10.1007/s10462-023-10517-0

Plant diseases impact extensively on agricultural production growth. It results in a price hike on food grains and vegetables. To reduce economic loss and to predict yield loss, early detection of plant disease is highly essential. Current plant disease detection involves the physical presence of domain experts to ascertain the disease; this approach has significant limitations, namely: domain experts need to move from one place to another place which involves transportation cost as well as travel time; heavy transportation charge makes the domain expert not travel a long distance, and domain experts may not be available all the time, and though the domain experts are available, the domain expert(s) may charge high consultation charge which may not be feasible for many farmers. Thus, there is a need for a cost-effective, robust automated plant disease detection or classification approach. In this line, various plant disease detection approaches are proposed in the literature. This systematic study provides various Deep Learning-based and Machine Learning-based plant disease detection or classification approaches; 160 diverse research works are considered in this study, which comprises single network models, hybrid models, and also real-time detection approaches. Around 57 studies considered multiple plants, and 103 works considered a single plant. 50 different plant leaf disease datasets are discussed, which include publicly available and publicly unavailable datasets. This study also discusses the various challenges and research gaps in plant disease detection. This study also highlighted the importance of hyperparameters in deep learning.