Deep learning-based damage detection on cherry leaves

In this study, the use of deep learning methods is aimed at detecting diseases in cherry leaves to increase agricultural productivity. Currently, the detection of leaf diseases is carried out by expert personnel, but it can be time-consuming. Additionally, the number of these experts may be insufficient, indicating the difficulty of accurate detection. Therefore, the main goal of this study is to utilize deep learning-based disease detection applications to increase cherry production and diagnose diseases early. The performance of different deep learning models for cherry leaf disease detection has been investigated on two different datasets in the study. According to the average results, the MobileNet-V2 model demonstrated high performance with 98.68% accuracy and 98.83% precision, the Inception-V3 model provided satisfactory results with a 94.17% F1-Score. The proposed CNN model attracted attention with a 95.83% F1-Score, the proposed CNN + LSTM model presented satisfactory results with a 96.33% recall, and the proposed CNN + BiLSTM model exhibited high performance with a 96.33% F1-Score. In the category of best results, the MobileNet-V2 model yielded the highest results with a 99.03% accuracy rate and a 99.33% F1-Score, the Inception-V3 model showed satisfactory performance with 94.00% accuracy and 94.33% precision. The proposed CNN model demonstrated good performance with a 95.66% F1-Score, the proposed CNN + LSTM model provided the best results with 96.67% accuracy and 96.67% F1-Score, and the proposed CNN + BiLSTM model exhibited high performance with a 96.66% F1-Score. These results indicate promising outcomes for the use of deep learning models in cherry leaf disease detection. The best results in the study were obtained by the MobileNet-V2 and the proposed CNN + LSTM models. The reliability of this study can be enhanced by using various datasets, disease detection success rates can be increased by employing different deep learning methods, and disease detection times can be reduced. Keywords: Deep Learning, CNN, LSTM, Transfer Learning, Cherry Leaf Diseases