Preserving traditional plant knowledge via AI-driven identification and human-centered mobile design

Abstract

The loss of indigenous botanical knowledge poses risks to biodiversity conservation and community health by weakening the knowledge of medicinal plants and their traditional use as local plant-based health care options. An integrated system that combines deep learning image recognition with a user-friendly mobile interface to document and preserve medicinal plant knowledge is presented in this study. A Convolutional neural network (CNN) was trained on a curated dataset of fourteen Nigerian medicinal plants, achieving 78.9% validation accuracy in species identification. Recognized plants are linked to an expert-curated knowledge base (with features, compounds, and uses) structured as a graph for efficient retrieval. A mobile app prototype was also designed with intuitive user interfaces, such as a landing page for navigation and an image-based search screen for plant identification. The Human-Computer Interaction (HCI)-driven design emphasizes ease-of-use, with clear labels and minimal steps, enabling users (including elders and youth) to capture plant images and instantly access scientific and local names, medicinal properties, and usage guidelines. In “active learning” mode, low-confidence predictions trigger the retrieval of similar examples to guide users. This Artificial Intelligence plus Human-Computer Interaction approach addresses issues of manual identification (subjectivity, expertise gap) by automating recognition and by empowering communities to digitally record ethnobotanical knowledge. By improving accessibility to reliable plant data, this system contributes to Sustainable Development Goals (SDG) 3 and 15.