AntID_APP: Empowering Citizen Scientists with YOLO Models for Ant Identification in Taiwan

Ants play a vital role in ecosystems by serving as bioindicators and contributing to soil health and food webs. Accurate ant identification is therefore crucial for biodiversity monitoring, ecological research, and conservation efforts, particularly in biodiverse regions like Taiwan. However, traditional taxonomic methods are often time-consuming and require specialized expertise, which limits large-scale data collection and public participation. This limitation underscores the need for accessible and scalable identification tools. This paper presents AntID_APP, a web-based application designed to democratize ant identification for citizen scientists in Taiwan (https://webs.cloud.ncnu.edu.tw). The application leverages real-time image analysis to identify 54 ant genera native to the region. The underlying models were trained using datasets sourced from the open-access, annotated iNaturalist database. At the core of the system is YOLO, a cutting-edge object detection model that we fine-tuned for the rapid and precise identification of ants from user-submitted photographs. The application features a user-friendly interface that allows individuals to easily upload images. These are then processed by the embedded YOLO model, which delivers efficient and reliable genus suggestions even when dealing with the variable image qualities typical of field observations. By integrating advanced AI, the system transforms complex taxonomic tasks into an engaging experience for non-experts. Our results demonstrate high identification accuracy, validating the tool's potential for large-scale citizen science initiatives. Furthermore, the combination of an intuitive web interface and a lightweight server architecture not only fosters public engagement in science but also provides a practical tool for advancing biodiversity monitoring and environmental conservation.