Optimization of models for recognizing diseased plants of tomato leaves using deep learning technologies

Abstract

In this work, using deep learning technologies as an element of artificial intelligence, various neural network architectures were built that model diseases of strategically important agricultural plants. Modern deep learning optimizers are proposed, and optimization problems for the proposed neural network architectures are studied. The main tasks of agriculture are identified, which today is a key sector for determining economic stability and solving food security. In this work, models for predicting yields based on deep learning are built. The main tasks of agriculture, which today are relevant in the context of climate change, are substantiated. The consequences and causes of the influence of climate change on many large categories of agriculture have been identified. The reasons for the influence of temperature anomalies in summer and winter on plant diseases have been established. The work examines voluminous concepts such as the yield of agricultural crops that lose stability with the rapid spread of various plant diseases, causing millions of losses to farmers and agricultural producers. Methods based on artificial intelligence have been proposed that can prevent the causes and factors of plant diseases. Models created on the basis of neural networks and optimizers are proposed to guarantee the accuracy of models for identifying and early diagnosing plant diseases. In the work, based on models built on the basis of deep learning, plant diseases are predicted using a database of sick and healthy plant images. The accuracy of the models has been established, which can guarantee high-quality prediction of the task. Based on the problem of optimizing neural networks for recognizing plant diseases, a mechanism for applying the selection of optimal parameters and selecting neural network architectures has been developed. The accuracy of the created models was analyzed, which is an assessment of their quality. When optimizing a neural network, the main attention in the work is paid to the task of choosing optimizers. Based on modern optimizers, various models have been obtained for predicting plant diseases based on a database collected in the field.