Citrus Leaf Scab Multi-Class Classification: A Hybrid Deep Learning Model for Precision Agriculture

Growing citrus fruit is crucial to the agriculture industry because of the economic benefits and consumer demand it provides across the world. However, citrus leaf scab illnesses brought on by the pathogenic fungus Elsino fawcettii remain a constant danger to the thriving citrus sector. Preventing crop losses and ensuring sustainable citrus output depends on early and precise identification of these diseases. The proposed study combines the accuracy of convolutional neural networks (CNN) with the transparency of decision tree (DT) to solve the problem of citrus leaf scab classification. The current work amassed a massive dataset of 50000 high-resolution photos of citrus leaves with varying degrees of citrus leaf scab disease to use in the analysis. Expert plant pathologists painstakingly labeled the dataset and split it into training, validation, and testing groups. Preprocessing methods were used to normalize the data and increase the size of the training set. Along with this, it presents a hybrid model that uses a CNN to analyze and make transparent decisions based on the pictures of citrus leaves. The model was trained and tested to determine how well it could classify citrus leaf scab samples into many categories. The total accuracy of the proposed approach in diagnosing the severity of citrus leaf scab illnesses was 98.07%. The accuracy and the capacity to manage unbalanced classes were further confirmed by the precision, specificity, and F-measure for each acuteness level (ACL). The suggested study contributes to the field of precision agriculture by suggesting an accurate and mechanised method for identifying citrus leaf scab. Through this study, the authors want to equip citrus farmers with a diagnostic tool that will allow for early disease diagnosis, leading to more effective and efficient treatments and ultimately, more sustainable citrus output.