AI-based identification of Chinese vascular plants from herbarium specimens: a tool for all herbaria with Chinese holdings.

Journal: The New phytologist
Published Date: Aug 9, 2025

Abstract

Documenting the geographic ranges of the World's > 350 000 named species of vascular plants requires artificial intelligence (AI) because there are insufficient experts to speedily identify dried plants, a task that requires access to thousands of already named specimens representing the applicable names. Focussing on herbarium material, we built an AI model that helps identify China's vascular plants. Training and testing sets consisted of images representing 21 589 species (including 1866 infraspecific taxa), 3144 genera, and 309 families of vascular plants represented in Chinese herbaria that have digitized over 6.5 million images of herbarium specimens identified to species level. The resulting AI model is able to identify all 21 589 species, covering an estimated 53.9% of China's vascular plant diversity. A test of the model with 244 749 expert-identified specimen images representing the species resulted in a species-level accuracy of 67.3% for Top-1 and 88.2% for Top-5. Our '21Ksp' model is freely accessible at https://www.iplant.cn/stu/spm. The new model, and a related tool with identified live photographs of 20 000 vascular species common in China (FlowerMate 2.0, https://www.iplant.cn/stu/20k), will help reduce the backlog of unidentified specimens in hundreds of herbaria world-wide that harbour collections from China and adjacent countries.

Authors

  • Gan Xie
    Big Data and AI Biodiversity Conservation Research Center, Institute of Botany, Chinese Academy of Sciences, 20 Nanxincun, Xiangshan, Beijing, 100093, China.
  • Jing Xuan
    College of Language Intelligence, Sichuan International Studies University, Chongqing 400067, China.
  • Hai-Rui Luo
    Big Data and AI Biodiversity Conservation Research Center, Institute of Botany, Chinese Academy of Sciences, 20 Nanxincun, Xiangshan, Beijing, 100093, China.
  • Hui-Yuan Liu
    National Plant Specimen Resource Center, Institute of Botany, Chinese Academy of Sciences, 20 Nanxincun, Xiangshan, Beijing, 100093, China.
  • Susanne S Renner
    Department of Biology, Washington University in Saint Louis, St Louis, MO, 63130, USA.
  • Min Li
    Hubei Provincial Institute for Food Supervision and Test, Hubei Provincial Engineering and Technology Research Center for Food Quality and Safety Test, Wuhan 430075, China.

Keywords

No keywords available for this article.

External Resources

Popular Topics
Recent Journals