Cloud-Assisted Deep Learning Framework for Real-Time Cotton Leaf Disease Diagnosis using Android Application

Abstract

Early and accurate detection of cotton leaf diseases is essential for improving crop productivity, minimizing yield losses, and supporting sustainable agricultural practices. Traditional disease diagnosis methods depend on manual field inspection by experts, which is time-consuming, subjective, and often impractical for large-scale cotton cultivation. To address these limitations, this research presents an intelligent Android-based cotton leaf disease classification system that integrates mobile image acquisition, cloud-based storage, and algorithm-driven disease classification.
In the proposed framework, cotton leaf images are captured or selected using an Android mobile application and securely uploaded to the cloud using the ImgBB image hosting service. The cloud platform provides scalable image storage and acts as an interface between the mobile application and the classification engine. Uploaded images are retrieved from the cloud and preprocessed through resizing, normalization, and enhancement to ensure consistency with the trained dataset. Disease classification is then performed using machine learning and deep learning algorithms, with emphasis on a transfer learning–based CNN–Transformer–LSTM hybrid model trained on cotton leaf images representing multiple disease classes and healthy leaves.
The classification module automatically identifies cotton leaf diseases such as Alternaria leaf spot, bacterial blight, and grey mildew, along with healthy leaf conditions, and returns the predicted class with confidence scores to the Android application. By offloading computationally intensive tasks to the cloud, the system minimizes mobile device resource usage while maintaining high classification accuracy. Experimental evaluation on the cotton leaf dataset demonstrates strong performance in terms of accuracy, precision, recall, and ROC–AUC values, confirming the robustness of the proposed approach under varying image capture conditions.
The developed Android–cloud integrated system provides a practical, user-friendly, and scalable solution for real-time cotton leaf disease diagnosis. It enables farmers and agricultural practitioners to obtain rapid diagnostic feedback using smartphones, facilitating early disease management and reducing dependence on expert intervention. The proposed framework can be extended to support additional crops, real-time model updates, and integration with IoT-based agricultural monitoring systems, making it a valuable contribution to smart and precision agriculture.