Intelligent Disease Detection in Mango Leaves: A CNN-SVM Approach

Mango leaf disease is a serious affliction affecting the quantity and quality of crops which in turn warrants the development of efficient detection techniques. This research project focuses on a deep learning-based classification model for the identification of six mango leaf diseases Tip, Phoma Blight, Dieback, Sooty Mold, Leaf Gall, and Red Rust. The proposed model has shown great potential in terms of classification accuracy with Precision values ranging from 95.53% for Sooty Mold to 96.71% for Dieback. Similarly, Recall values varied between 95.74% for Black Tip to 96.53% for Phoma Blight while F1-Score ranged from 95.80% for Sooty Mold to 96.49% for Dieback, proving the performance of the model in the identification of infected leaves. The model was validated using a test set of 11,580 images for reliability and strength. Support Ratio for the various classes of disease ranged from 0.15 (Sooty Mold) to 0.18 (Dieback), yielding a very good overall accuracy of 99% for all classes combined. The results justify the efficacy of the model in identifying infected leaves and provide a reliable method for early detection of disease. The integration of this model into computerized farm monitoring systems would allow the farmers to optimize disease control measures for timely intervention and healthier crops. The accuracy and precision of this technique make it a strong candidate for sustainable mango cultivation, leading to improved productivity and reduced losses owing to diseases.