Smart farming for detection and identification of tomato plant diseases using light weight deep neural network

• Published in: Multimedia tools and applications Vol. 82; no. 12; pp. 18799 - 18810
• Main Authors: Aishwarya, N., Praveena, N. G., Priyanka, S., Pramod, J.
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.05.2023; Springer Nature B.V
• Subjects: Artificial neural networks; Classification; Computer Communication Networks; Computer Science; Computing costs; Crop diseases; Crop production; Data Structures and Information Theory; Machine learning; Model accuracy; Multimedia Information Systems; Neural networks; Plant diseases; Smartphones; Special Purpose and Application-Based Systems; Tomatoes; Deep learning; Smart farming; Convolution neural network; Plant disease
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-022-14272-2

Tomato occupies a protuberant place in every kitchen in various forms and this is one of the finest crops cultivated worldwide. However, the production of tomato crops is adversely affected due to different type of diseases. On one hand, timely recognition and classification of diseases is very essential for maintaining both the quality and quantity of tomatoes. And on the other hand, the combination of rising worldwide smartphone adoption and recent deep learning advancements has paved the way for smartphone-assisted disease detection. Hence, to accomplish the above-mentioned perspectives, a custom CNN model (CCNN) is designed to classify various tomato plant diseases. In contrast to state-of-the-art architectures such as Alex net and VGG-16, the proposed CCNN model has three convolution layers followed by three fully connected layers which reduces the processing time and the computational power while achieving a greater accuracy in classification of various diseases. Also, the number of hyper parameters of the proposed model is significantly decreased as compared with the existing models. The competency of the proposed model is experimentally verified with 10 classes of tomato leaves both qualitatively and quantitatively. The results exhibit that the Custom CNN model attains competitive accuracy as compared with the conventional models with lesser computational cost. Further, the deployment of the proposed CCNN model in the mobile based system paves way for widespread global smartphone-assisted crop disease diagnosis.