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

Area of Science:

• Ecology
• Bioinformatics
• Conservation Biology

Background:

• Ants are crucial bioindicators for soil health and food webs, essential for biodiversity monitoring.
• Traditional ant identification methods are labor-intensive and require expert knowledge, hindering large-scale data collection.

Purpose of the Study:

• To develop AntID_APP, a web application for real-time, image-based detection and genus-level identification of native ants in Taiwan.
• To empower citizen scientists with an accessible tool for ecological research and public engagement.

Main Methods:

• Utilized fine-tuned YOLO (You Only Look Once) models for ant detection and genus classification.
• Trained models on a dataset of 60,429 iNaturalist images covering 54 native ant species.
• Implemented targeted data augmentation and evaluated multiple YOLO versions (v9-v12) for optimal performance.

Main Results:

• Achieved high performance in ant detection with mean Average Precision (mAP50: 0.935-0.948, mAP50-95: 0.777-0.807).
• Demonstrated accurate genus-level identification of ants from user-uploaded images.
• Developed an intuitive interface with a lightweight asynchronous server for efficient image processing and results delivery.

Conclusions:

• AntID_APP provides a scalable and accessible solution for biodiversity monitoring through citizen science.
• The application enhances public engagement in ecological research by simplifying ant identification.
• Facilitates accurate, real-time data collection for ant genus identification in Taiwan.