Hyeonji Moonp1, Su-Jeong Han2, Jaesung Lee2, Sangtae Kim1

1Department of Biology, Sungshin Women's University, Seoul, Rep. of Korea;

2Department of Artificial Intelligence, Chung-Ang University, Seoul, Rep. of Korea

Invasive alien plants are non-native species that threaten biodiversity by exhibiting rapid growth and reproduction beyond their native ranges. Timely monitoring is essential for effective management, yet expert surveys are costly and time-consuming. Citizen science platforms, such as iNaturalist and NATURING, offer large-scale monitoring opportunities; however, accurate identification is challenging without taxonomic expertise, especially during non-flowering stages. We developed a deep learning model to identify invasive alien plants in riparian habitats of the Korean Peninsula using citizen science images. The model was trained on nine invasive alien species and 21 morphologically similar native or non-target species. A curated dataset from NATURING and iNaturalist covered variation across growth stages and organs. We applied ResNet, a deep learning model, and evaluated its performance using F-score, accuracy, and confusion matrices. This study shows the potential of AI-assisted tools to enhance citizen science–based monitoring and education on invasive alien plants, highlighting the research value of citizen science data and improving identification reliability.