Medicinal plant recognition based on Vision Transformer and BEiT
Duy Tran Nguyen Nhut, Thinh Duong Tan, Trung Nguyen Quoc, Vinh Truong Hoang
Abstract
The use of deep learning models for plant identification has garnered significant attention in the research community and has yielded promising results. In this study, we evaluate the performance of four state-of-the-art pre-trained deep learning models, namely EfficientNetBO, EfficientNetV2-S, Vision Transformer (ViT), and Bidirectional Encoder Image Transformer (BEiT), on the VNPlant-200 dataset, a complex dataset that comprises various species of medicinal plants captured in natural settings. Our results show that BEiT achieved the highest accuracy of 99.14%, outperforming the other models evaluated on this benchmark. These findings prove the effectiveness of these models in plant recognition tasks, particularly in the context of medicinal plants.