Detection of Tomato leaf diseases using Attention Embedded Hyper-parameter Learning Optimization in CNN

India’s agricultural productivity is vital, as its economy depends heavily on it. Plant disease identification in plants is significant in the agricultural sector, as a disease in plants is very natural. In this setting, plants are susceptible to damage that reduces their quality, quantity, or production if proper precautions are not followed. Tomatoes are the most common crop globally, and they can be used in various ways in any kitchen, regardless of cuisine. It is the third most widely grown crop globally, after potatoes and sweet potatoes. India was ranked second in tomato production. However, the quality and quantity of tomato crops suffer from numerous diseases. As a result, the paper discusses a deep learning-based approach to disease detection. Outdated approaches focus on handcrafted features extracted from obtained images to regulate infection. The quality of the handmade elements selected is also crucial to the success of these works. This problem may be handled by using Convolutional Neural Networks for automated feature learning (CNN). The research presented here illustrates two different methods for identifying infected tomato leaves. Hyper-parameter learning with an optimization technique is employed first to study the important features, and the second design employs an attention mechanism. Finally, this model is tested by identifying three diseases–leaf mould, late blight, and early diseases–and classifying them in a publicly accessible dataset called Plant Village Dataset. Data augmentation has to be created as future work to increase categorization accuracy.