A Cloud-Enabled Deep Learning Framework for AI Driven Plant Disease Detection

Timely diagnosis of plant disease is important in avoiding declines in crops and ensuring food security. To facilitate real-time detection of plant diseases, we present a cloud-based mobile diagnostic system through deep learning (DL) techniques. It allows users to click or upload plant leaf images via a specific smartphone app. The system adopts a hybrid CNN-VGG16 strategy, with protected image transmission via Ngrok to the cloud server and delivers a 99.54% accuracy in the classification of plant diseases, thus outperforming the conventional models of CNN+LSTM <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{(9 5. 3 \%)}$</tex>, LSTM (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{8 0. 6 \%}$</tex>), and Logistic Regression (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{7 5. 1 \%}$</tex>). The method also significantly reduces the computation load on edge devices while supporting compatible deployment on the cloud platform. Robustness of the model is demonstrated through measures like sensitivity, accuracy, and F1-score. Through automated diagnostic capabilities, the solution effectively bridges AI research to real-world applications in precision agriculture, thus improving farmers' capacities to respond rapidly to potential risks.