<![CDATA[The rapid advancement of Artificial Intelligence (AI) technology has created new opportunities for modernizing the agricultural sector, particularly in the early detection and classification of plant diseases based on digital images. A Computer Vision-based approach has emerged as an effective solution, as it enables the automation of visual analysis that was previously reliant on manual observation. In this study, a method based on Fusion Convolutional Neural Networks (CNN) is proposed, combining the strengths of ResNet and DenseNet architectures to produce more robust and discriminative feature representations. In addition, this research integrates an Explainable Artificial Intelligence (XAI) technique using Grad-CAM to provide visual interpretations of the model’s decisions, thereby enhancing user trust in the developed system. The dataset used consists of three main classes of chili leaf conditions: Bacterial Spot, Curl Virus, and Healthy. Experimental results demonstrate that the proposed model achieves excellent performance, with an accuracy of 98%. Further analysis through the classification report indicates that the Healthy class attains perfect performance, with precision, recall, and f1-score all reaching 1.00. Meanwhile, the Bacterial Spot class achieves a recall of 1.00 and an f1-score of 0.97, indicating the model’s capability to correctly identify all samples in this class. The Curl Virus class also shows strong performance, with a precision of 1.00, recall of 0.95, and f1-score of 0.97. Overall, the macro average and weighted average f1-scores both reach 0.98, reflecting the model’s stability and consistency across all classes. Furthermore, the implementation of Grad-CAM is able to highlight specific regions on chili leaves that contribute to the model’s predictions, providing deeper insight into the disease patterns recognized by the model. This not only enhances interpretability but also supports visual validation by users. Therefore, this study demonstrates that the combination of Fusion CNN and Grad-CAM is not only effective in improving classification accuracy but also ensures transparency in the decision-making process, making it highly suitable for intelligent decision-support systems in precision agriculture]]>
Copyrights © 2026
