Deep Learning-Based Multiclass Classification of Diseases in Cucumber Fruit: Enhancing Agriculture Diagnosis

Agriculture is central to the economy of many developing countries.Cucumber (cucurbits) is the most widely grown commercial vegetable in terms of the production, the production is greatly affected by diseases and the yield loss of cucumber plant is greatly determined by disease such as Belly Rot and Pythium fruit rot.Early and reliable disease diagnosis is crucial for effective controlling to reduce the economic loss caused by the diseases and improve the quality of crops.Conventional methods are visual identification-based and time consuming, and also not accurate, especially in early stages of the disease.In this work, we attempt to address these issues and propose an automated cucumber disease classification system using transfer learning.A retraining was performed on three convolutional neural network models (VGG16 pre trained, MobileNetV2 and ResNet-50) using a database of 2400 images with two disease classes and a normal one.The images were preprocessed by Contrast Limited Adaptive Histogram Equalization (CLAHE) and background removal by deep learning segmentation to clear the background noise and concentrate on the informative feature of the image.The models were trained and tested using training, validation, and test sets, and achieved the accuracy of 95.28%, 98.06% and 57.5%, respectively.MobileNetV2 achieved the best performance among all models, such as the highest precision, recall, and F1 score of 0.98, showing the robustness and real-time disease classification ability.The experimental results indicate that transfer learning can effectively solve the problem of insufficient labeled cumber diseased leaf samples and the difference of acquisition and intensity of diseased leaf images, and is a concrete model for early detection of diseases for cucumbers.The proposed system has potential to aid the farmer and agronomist in early disease management decisions and cut on chemical application.In the future, we will continue to add more disease class into the data set to improve it, and develop a mobile APP for disease detection.