IoT-Integrated Deep Learning Framework for Real-Time Plant Leaf Disease Detection in Precision Agriculture

Diseases have a tremendous effect on agriculture and food security of the world and the economy. Early and precise detection of leaf diseases is crucial for minimising any losses in the crop, optimal use of pesticides and other safe farming practices. This study aspires to develop an IoT integrated framework of Deep learning with the aim of real time detection of disease in the leaves for the context of precision farming. The proposed system utilizes the Convolutional Neural Networks (CNNs) and transfer learning technique for automatic removal of the discriminative features from the leaves images captured in real world environment. The incorporation of the spatial attention mechanism into the model, which helps the model to identify the affected area of the disease, makes the model interpretable, and the optimization methods such as pruning and quantization make the model light and convenient to deploy on edge devices. The system is composed of three layers namely a perception layer comprising IoT- processing the image, the network layer involved in deep learning-based inference using cloud and edge computing, and finally the application layer involved in providing actionable insights through the decision support dashboards. While the conventional clean data datasets can be used to improve generalization and practicality the same goes for the noisy datasets in the field. The system supports real time classification of diseases, estimation of severity and geospatial monitoring which helps to take precision agriculture to scale. The experimental analysis shows that the proposed architecture gains high classification accuracy, remarkable computation efficiency and good adaptability under various agricultural conditions. Overall, the findings of this study contribute to deep learning for agricultural disease detection and propel the closer adoption of machine learning in agrostech applications, facilitating intelligent, scalable and sustainable disease management in the realm of gardening.