Deep Learning Meets Support Vector Machines: An Effective Hybrid Model for Banana Leaf Wilt Disease Severity Assessment

Throughout the world, bananas are grown as a staple crop that contributes significantly to the economy. However, diseases like banana leaf wilt pose serious hazards that, if left unchecked, might lead to devastating crop losses. In this study, we introduce a fresh method for dealing with this issue by creating a model that combines features of both a Convolutional Neural Network (CNN) and a Support Vector Machine (SVM). Our model uses deep learning (DL) to categorize the severity of banana leaf wilt, empowering farmers to take preventative measures as soon as the disease is detected. We started by compiling a large collection of banana leaf photos, which were then labeled by domain experts according to severity. The quality and diversity of the dataset were guaranteed by employing data pretreatment and augmentation strategies. As the feature extractor, the selected pre-trained CNN architecture captured subtle patterns that are highly predictive of illness severity. In addition to a remarkable 94.77% overall accuracy, our hybrid model also offered a sophisticated evaluation of illness development. To provide farmers with accurate insights regarding the state of their crops, performance parameters such as precision, recall, and Fl-score were calculated for each severity level. In addition, a comparison with state-of-the-art models in the field confirmed the model's superiority. Its potential as a game-changing tool for illness severity classification was reaffirmed by its repeated outperformance of previous techniques.