Implementation of an EfficientNet-B4 Model Architecture with a Convolutional Block Attention Module (CBAM) for Betel Leaf Disease Classification

Betel leaf farming plays a significant role in the agricultural economy of Southeast Asia, particularly in Indonesia, supporting the cultural practices and rural livelihoods. However, the sector faces challenges from diseases caused by fungal, bacterial, and viral pathogens, leading to significant yield losses. Traditional leaf disease detection methods are limited in accuracy and timeliness, necessitating innovative solutions. This study presents an advanced approach leveraging EfficientNet-B4, enhanced with the Convolutional Block Attention Module (CBAM), for betel leaf disease detection. A localized dataset of 4,000 high-resolution images of betel leaves, categorized into four classes, was used to ensure relevance to the Indonesian agriculture. The CBAM-enhanced model demonstrated superior performance in identifying disease-specific patterns, achieving an average accuracy of 95.6%, compared to the 90% with the base model. Using metrics, such as precision, recall, F1-score, and ROC-AUC, the proposed methodology highlights its robustness and reliability. The study’s findings underscore the importance of localized datasets and attention mechanisms in improving the disease classification accuracy. Practical implications include the potential for real-time deployment of the model on mobile platforms, enabling early detection and intervention. This approach promises to enhance the sustainability of betel leaf farming by minimizing the crop losses, reducing chemical usage, and supporting the farmers’ economic well-being.