Rotula Aquatica Medicinal Plant Detection using ViT with Explainable AI

This research paper presents a system for recognizing medicinal plants using Vision Transformer (ViT) technology and an AI Explained Module. In particular, its purpose is for the purpose of automatically recognizing and explaining Rotula aquatica - which is a rare, medicinal plant and known colloquially as water pepper. The system blends a transformer based deep learning model that identifies a plant from an image while providing a level of contextual information that includes taxonomical classification, the habitat in which this plant can be found, and medicinal characteristics. The ViT model was trained on a dataset of images of medicinal plants and optimized for use on a mobile device using TensorFlow Lite. The AI Explained module will automatically identify if it's a Rotula aquatica and present a generated response with various levels of detail to the user. Experimental results indicated that the ViT model had a reported accuracy of 94.8% as an automatic identification tool that outperforms traditional CNN models. Our system provides a connection of knowledge presentation with AI-based recognition to support an education and conservation efforts.