Citrus Leaf Disease Detection Using Hybrid CNN-RF Model

Abstract

One of the nutrient-dense foods with anti-oxidant and anti-mutagenic are citrus fruits. The protection of citrus fruit and leaves against infectious diseases is very essential. Diseases of citrus fruits are the primary factor contributing to the drastic reduction in citrus fruit yields. Therefore, it is important to develop a system that can automatically detect and identify citrus leaf diseases. Deep learning algorithms have shown success in Artificial Intelligence (AI) challenges in recent years. The proposed model is a hybrid of machine learning and a convolutional neural network (CNN). In order to identify citrus leaf diseases, this research uses a CNN model as a feature extractor and a Random forest for classification. The model has been trained and tested using 58 healthy and 536 unhealthy images of Black Spot, Canker and Greening. The best result was obtained by the VGG16-Random Forest algorithm, which had an accuracy of 87%. ResNet50-Random Forest obtained 83% accuracy, whereas InceptionV3-Random Forest obtained the least accuracy with 80% out of the three.