IoT and CNN-based Tomato Leaf Disease Monitoring and Detection with Android Application

Abstract

Tomatoes are a high-value commodity in agriculture, so farmers make various efforts to ensure the production of fresh and ready-to-consume tomatoes. However, farmers often face difficulties in monitoring tomato growth because they still use manual methods and have limited knowledge in detecting diseases on tomato leaves. This research offers a solution by utilizing transfer learning and fine-tuning Convolutional Neural Network (CNN) using DenseNet169 architecture, as well as Internet of Things (IoT) technology. The model is implemented in an Android application using TensorFlow on the Flutter platform after being converted to tflite format. The test results show that the accuracy of the model reaches 94%, while the accuracy of the application in detecting tomato leaf diseases reaches 92.80% and has a response time of about 1077.56 ms. In addition, the application can monitor plant conditions in realtime by having a delay of 1,998 ms.