A Lightweight Deep Learning Model for Crop Disease Detection on Mobile Devices

Early detection of crop diseases plays a vital role in modern agriculture by reducing yield losses and enhancing food security. This study aims to develop and evaluate lightweight deep learning–based models for crop disease detection, with a focus on deployment on mobile devices. Using simulated agricultural data incorporating soil health, weather conditions, and plant health parameters, three machine learning models—random forest (RF), support vector machine (SVM), and convolutional neural network (CNN)—were trained and tested. Model performance was evaluated using accuracy, F1-score, and inference time. Experimental results within the simulated dataset show that RF and SVM achieved perfect classification performance (100% accuracy and F1-score of 1.0). These results demonstrate the effectiveness of the models under controlled synthetic conditions, while further validation using real-world agricultural data is required. The CNN achieved 87.5% accuracy with a loss of 0.2279. Despite lower accuracy, the CNN demonstrates strong potential for real-time mobile applications. The study highlights opportunities for optimizing lightweight CNN architectures using real-world agricultural data in future work.