Abstract:
This study presents the development of an AI-driven tool for detecting bacterial
and fungal diseases in jackfruit using advanced image processing and deep
learning techniques. Leveraging four pretrained convolutional neural network
(CNN) models VGG19, MobileNetV2, EfficientNetB0, and ResNet50, alongside
an ensemble approach, the study focuses on achieving high accuracy and
reliability in disease classification. The dataset, comprising jackfruit images
categorized into Healthy, Bacteria Affected, and Fungus Affected classes,
undergoes preprocessing steps such as normalization, resizing, and
augmentation to ensure robust training and evaluation. Among the individual
CNN models, VGG19 and MobileNetV2 demonstrate superior performance, with
accuracies of 95.84% and 93.35% on the test set, respectively, while
EfficientNetB0 exhibits the lowest performance due to instability in learning.
The ensemble model significantly enhances classification performance,
achieving a near-perfect accuracy of 99.83% and an AUC score of 1.00 across all
classes, combining the strengths of individual models to minimize
misclassifications. Furthermore, the study implements the model in a mobile
application, "Jackfruit-Doctor," providing real-time disease detection with high
accuracy and accessibility for end users. This application empowers farmers by
enabling early intervention, reducing crop losses, and promoting sustainable
agricultural practices. The findings demonstrate the efficacy of integrating
ensemble learning with mobile technology, offering a reliable and scalable
solution for agricultural disease management while contributing to food security
and economic sustainability.
