AlexNet Convolutional Neural Network for Disease Detection and Classification of Tomato Leaf

With limited retrieval of reserves and restricted capability in plant pathology, automation of processes becomes essential. All over the world, farmers are struggling to prevent various harm from bacteria or pathogens such as viruses, fungi, worms, protozoa, and insects. Deep learning is currently widely used across a wide range of applications, including desktop, web, and mobile. In this study, the authors attempt to implement the function of AlexNet modification architecture-based CNN on the Android platform to predict tomato diseases based on leaf image. A dataset with of 18,345 training data and 4,585 testing data was used to create the predictive model. The information is separated into ten labels for tomato leaf diseases, each with 64 × 64 RGB pixels. The best model using the Adam optimizer with a realizing rate of 0.0005, the number of epochs 75, batch size 128, and an uncompromising cross-entropy loss function, has a high model accuracy with an average of 98%, a strictness rate of 0.98, a recall value of 0.99, and an F1-count of 0.98 with a loss of 0.1331, so that the classification results are good and very precise.