Artificial intelligence-based heritage tree disease diagnosis using transfer learning: A case study of Zelkova serrata

다국어 초록 (Multilingual Abstract)

Heritage trees are highly valued and protected by national laws because of their cultural and historical significance. However, due to the financial and time constraints associated with the continuous monitoring of the heritage trees, unnecessary losses are reported annually. As a solution, this study suggests an artificial intelligence(AI)-based heritage tree disease diagnosis system on Zelkova serrata. We have compared several state-of-the-art deep learning models with transfer learning on the Zelkova Serrata Dataset which consists of 680 images. All models achieved outstanding classification results even only using pre-trained weights of ImageNet, with F1 scores ranging from 92.00% to 96.26%. Particularly when additionally leveraging the plant disease datasets, model performances improved to a range of 93.78% to 99.45%. Through this research, we proposed the concept of AI-based heritage tree disease diagnosis using transfer learning. This system is expected to reduce the aforementioned financial and time constraints.