Ensemble Learning for Enhanced Mango Leaf Disease Classification: A Deep Learning Approach

Mangoes are highly consumed fruits especially in the tropical as well as subtropical zones of the earth. It is said that nearly half of the earth's population consumes mangoes in some form or the other either as fresh mangoes or in processed form such as juice, dried mangos, or other cooked dishes. That said, this significant food crop has a number of limitations in its cultivation and the least of these is attack by pests. The agro-ecosystem of mango plantation includes several factors such as soil, insects, diseases, weather and many more which determine the well-being of the mango trees in that plantation. Of all these concerns, leaf diseases turn out to be the most offensive by causing losses in both the quantity and quality of mangoes produced. To mitigate these problems, research has been done and continues to be done towards identifying and classifying the mango leaf diseases with better yields and less losses thanks to processing techniques such as deep learning and image processing. Recently published papers focus on these modern methods and analyze the advantages of their applications for the diagnosis and treatment of mango plants. This paper proposes a novel framework that identifies the diseases in mango leaves. The approach demonstrates an impressive 99.94% testing accuracy employing ensemble learning and transfer learning techniques. It is superior to other popular deep learning-based models that include DenseNet121, ResNet50 and VGG19. The results of the study demonstrate the efficiency of the proposed approach as well as its superiority in important performance metrics such as precision and recall. This evolution provides a viable answer to the very pertinent challenge of disease detection in mango cultivation allowing for better management of crops and hence increased production.