Mango leaf disease detections using deep learning models

Abstract

The detection and classification of diseases affecting mango leaves are critical for maintaining crop health and ensuring optimal yield. Manual inspection methods are often time-consuming and prone to human error, necessitating the development of automated detection systems. In this study, I propose a novel approach utilizing image data analysis for the detection of mango leaf diseases. Leveraging advanced image processing techniques and convolutional neural networks (CNNs), our model can accurately classify various types of leaf diseases based on image features. Our study integrates advanced techniques such as EfficientNetB4, VGG19, VGG16, InceptionV3 and Xception. By extracting relevant features from mango leaf images and training the model on a diverse dataset, we achieve robust performance in disease detection. Additionally, we explore the integration of transfer learning to enhance the model's capability to generalize across different disease types and environmental conditions. Through rigorous experimentation and validation, our framework demonstrates promising results, offering a reliable tool for early disease diagnosis and effective management strategies in mango cultivation. This research contributes to the advancement of precision agriculture practices, facilitating timely interventions and ultimately improving crop health and yield. The study tested integrated models for image classification, revealing that VGG19, InceptionV3, VGG16, EfficientNetB4, Xception and our proposed model demonstrated exceptional performance. Our proposed model has the height accuracy at 98%, while VGG19, InceptionV3, VGG16 and EfficientNetB4 secured the top positions with 86.50%, 96.49%, 89.99% and 94.09% accuracy. Transfer learning models improved accuracy, but remained lower than proposed models. This research is significant for data scientists as it highlights the importance of continuous advancements in agricultural research.