AI-Powered Herbal Knowledge Preservation & Identification System

Abstract

Medicinal plants have always been vital to human health and remain central to traditional and modern medicine. The World Health Organization (WHO) reports that about 80% of people in developing nations depend on herbal remedies for primary healthcare. However, identifying these plants manually is complex, time-consuming, and requires expert botanists, making it inaccessible to the general population. Furthermore, much of the traditional herbal wisdom is being lost because of poor documentation and modernization.
This paper presents an AI-Powered Herbal Knowledge Preservation and Identification System, a complete platform that combines deep-learning-based plant recognition with digital knowledge management. The proposed system employs Convolutional Neural Networks (CNNs) and transfer-learning models such as ResNet and EfficientNet for image classification, along with a collaborative herbal database that stores verified ethnomedicinal information. Large datasets are collected, preprocessed, and trained using these architectures to achieve high identification accuracy. Experimental evaluation shows a recognition accuracy of 92% (top-1) and 97% (top-5), which surpasses existing approaches like HerbVision and IndoHerb by 5–7%. Beyond classification, the system also enables community-based sharing of herbal information, ensuring that valuable ethnobotanical knowledge is preserved for future research, healthcare, and education.