Arvina Rizqi Nurul'aini, Mohammad Mahruf Alam, Rizky Ajie Aprilianto, Feddy Setio Pribadi
Plant damage due to disease is a serious challenge in the global agricultural sector, including onions (Allium cepa), whose production continues to decline due to infection with various leaf diseases. This work proposed a VGG16-ResNet50 hybrid model to detect onion leaf diseases through digital image processing. The dataset contains 4, 5 0 2 images of onions with various diseased and healthy conditions, which are processed through training, validation, and testing stages. The images are converted to 224 × 224 pixels to fit the VGG16 and ResNet50 architectures. The proposed model is compared to VGG16 and ResNet50 using evaluation metrics such as accuracy, precision, recall, F 1-score, and confusion matrix. The results show that the proposed hybrid model achieved the best performance, with an accuracy of 90.00%, precision of 91.11%, recall of 91.10%, and an F1-score of 90.90%, outperforming the individual VGG16 with accuracy 75.54%, precision 77.88%, recall 76.94%, F1-score 76.79% and ResNet50 with accuracy 88.97%, precision 89.97%, recall 90.01%, F1-score 89.81%. These findings demonstrate that the proposed hybrid model is more robust in accurately detecting onion diseases and has strong potential for implementation in artificial intelligence-based plant disease diagnosis systems within the precision agriculture sector. © 2025 IEEE.
Universitas Negeri Semarang, Department of Electrical Engineering, Semarang, Indonesia
