Ai-powered Indian Medicinal Plant Identification And Information System Using Deep Learning And Computer Vision Technology

Abstract

Good Identification Of Medicinal Plants Is The Key To Maintaining The Traditional Knowledge Systems And Safe Use Of Herbs. Physical Identification Of Plants Is Time Consuming And Subject To Errors Especially Where Species Have The Same Morphology. The Present Paper Describes An AI-based Indian Medicinal Plant Identification And Information System That Combines Convolutional Neural Networks With A Knowledge Retrieval System To Allow Real-time Identification Of Plants And Provide Surrounding Information About It. The Model Was Trained And Evaluated On A Curated Dataset Of 3,000 Images Of 50 Indian Medicinal Plant Species. Transfer Learning Was Used To Implement Two Deep Learning Architectures, ResNet18, And EfficientNet-B0. The Experimental Data Show That EfficientNet-B0 Was Able To Attain An Accuracy Of 90.3 Percent With A Latency Of 175 Ms To Run Inference, Surpassing Both ResNet18 And The Available Identification Systems (LeafSnap And PlantNet) To Identify Plants In Controlled Conditions. The System Also Includes Knowledge Base With Structured Knowledge Base And Language Model-based Retrieval System To Produce Detailed Information Such As Medicinal Uses, Phytochemical Properties, And Precautionary Advice To The Identified Species. Application Deployment Via A Streamlit Interface And Containerized Cloud Hosting Also Allows It To Scale And Be Accessed In Real Time. The Suggested Framework Illustrates The Usage Of The Computer Vision And Language-based Intelligence In An Efficient Manner, To Aid In The Field-level Medicinal Plant Detection And Digital Conservation Of The Herbal Knowledge.

Keywords