Deep Learning-Based Leaf Disease Detection with Mobile Application Integration for Smart Agriculture

Abstract

Today, agriculture is playing an important role in the food security and economic stability, crop diseases are still a huge problem in the world of crops and there is a great loss of yield. Plant disease detection is very important to reduce the agricultural losses and increase productivity. This paper is aimed at presenting a deep learning system for the detection of leaf diseases coupled with a mobile application for smart agriculture. The proposed system is based on the MobileNetV2 convolutional neural network architecture to classify the plant leaf diseases from the PlantVillage dataset of 20638 images into 15 different classes. To enhance model performance and generalization, techniques like resizing, normalization, augmentation, and shuffling data are implemented in the image preprocessing stage. The trained model is then converted to the TensorFlow Lite format and integrated into a mobile app developed using the Flutter framework, allowing for a mobile app to detect the diseases in real time on mobile devices without internet connectivity. Experiments show that the proposed method has superior classification performance and fast inference time, which is convenient for handheld devices.