Tomato Plant Disease Detection and Classification Using Convolutional Neural Network Architectures Technologies

Agriculture is efficient from an economic and industrial point of view. The majority of countries are trying to be self-sufficient to be able to feed their people. But unfortunately, several states are suffering enormously and are unable to join the standing up to satisfy their populations in sufficient quantities. Despite technological advances in scientific research and advances in genetics to improve the quality and quantity of agricultural products, today we find people who die of death. In addition to famines caused by wars and ethnic conflicts and above all plant diseases that can devastate entire crops and have harmful consequences for agricultural production. With the advancement of artificial intelligence and vision from computers, solutions have brought to many problems. Smartphone applications based on deep learning using convolutionary neural network for deep learning can detect and classify plant diseases according to their types. Thanks to these processes, many farmers have solved their harvesting problems (plant diseases) and considerably improved their yield and the quality of the harvest. In our article, we propose to study the plant disease (tomato) using the Plant Village [1] database with 18,162 images for 9 diseased classes and one seine class. The use of CNN architectures DenseNet169 [2] and Inception V3 [3] made it possible to detect and classify the various diseases of the tomato plant. We used transfer learning technology with a batch-size of 32 as well as the RMSprop and Adam optimizers. We, therefore, opted for a range of 80% for learning and 20% for the test with a period number of 100. We evaluated our results based on five criteria (number of parameters, top accuracy, accuracy, top loss, score) with an accuracy of 100%.