Utilizing ResNet Architectures for Identification of Tomato Diseases

Abstract

In this study, the effectiveness of various deep learning models in detecting diseases in tomato leaves has been thoroughly evaluated, with a particular emphasis on the ResNet architecture family. The primary focus was on assessing the capabilities of advanced models like Res2Next50 and Res2Net50d, which have shown exceptional performance in identifying and classifying various tomato leaf diseases. The results of the study reveal that the Res2Next50 model achieved the highest accuracy rate of 99.85%, accompanied by an impressive F1 score of 99.82%. This indicates a high level of precision and robustness in distinguishing between healthy and diseased leaves. Additionally, the Res2Net50d model has also proven to be highly effective, attaining an accuracy rate of 99.78%, precision of 99.73%, and recall of 99.72%. These metrics suggest that the model is not only capable of correctly identifying diseased leaves but also has a low false positive rate, making it a reliable tool for practical applications in agriculture. When compared to other popular convolutional neural network (CNN) architectures, such as VGG16 and DenseNet121, the ResNet models, particularly Res2Next50 and Res2Net50d, have demonstrated superior performance. This highlights the strength of the ResNet family in handling complex classification tasks, especially those involving fine-grained distinctions between different types of tomato leaf diseases. The findings of this research suggest that CNN models like Res2Next50 and Res2Net50d can significantly enhance the accuracy and reliability of automated disease detection systems in agricultural settings. Such systems can provide valuable support to farmers and agricultural professionals, enabling early and accurate identification of diseases and, consequently, better crop management and productivity.