An Improvement of Apple Leaf Diseases Detection Using Convolutional Neural Network Methods Based on Mobile Systems

One of the main problems in apple cultivation is leaf disease, which can reduce fruit production and quality. Some leaf diseases that commonly occur in apple plants are scab, cedar rust, black rot, and others. Various pathogens such as fungi, bacteria, and viruses can cause this disease and can spread through environmental factors and inappropriate cultivation practices. Treatment of foliar diseases in apple crops relies heavily on early diagnosis, effective pest and disease control, and the use of disease-resistant varieties. Therefore, early detection of apple leaf disease can help prevent other apple leaf diseases. This research uses a Convolutional neural network (CNN) algorithm with the Efficientnet-B7 architecture which is applied to a mobile system that uses Flutter technology to detect apple leaf diseases. With 8000 data, the results of this study are accurate in distinguishing or detecting apple leaf disease. After conducting several tests on various configurations, the most accurate results were loss 0.1007, accuracy 0.9642, validation-loss 0.0325, and accuracy value 0.9900. Apart from that, the f1 score for the accuracy, precision, recall, and confusion matrix stages is 0.99, or 99%.