SolenopsisDetector: development of an automatic detection system for fire ants using computer vision and
Song-Quan Ong1, Abdul Hafiz Ab Majid2, Wei-Jun Li3
1Institute for Tropical Biology and Conservation (ITBC), Universiti Malaysia Sabah, Kota Kinabalu, Sabah, Malaysia.
Journal of Economic Entomology
|August 4, 2026
Summary
An automated system, SolenopsisDetector (SolenopD), uses deep learning to identify invasive fire ants (Solenopsis spp.). It combines precise body segment localization with accurate classification, aiding rapid identification and management.
Area of Science:
- Entomology
- Computer Science
- Machine Learning
Background:
- Invasive fire ants (Solenopsis spp.) present significant ecological and economic challenges.
- Current identification relies on expert taxonomy, often leading to delays.
- Automated identification systems are needed to support pest management.
Purpose of the Study:
- To develop and evaluate SolenopsisDetector (SolenopD), an automated system for identifying Solenopsis ants.
- To compare whole-body versus segment-based image classification strategies.
- To assess the performance of deep learning models for ant detection and classification.
Main Methods:
- Trained and compared YOLOv5, YOLOv8, and YOLOv11 for ant and body segment detection.
- Evaluated ResNet, MobileNet, and InceptionV3 classification models using whole-body and segment-based images.
- Utilized 8,300 images for system development and validated with Grad-CAM for interpretability.
Main Results:
- YOLOv8 and YOLOv11 achieved superior detection performance (0.931 mAP for whole-body, 0.788 mAP for segment-based).
- Segment-based classification, particularly thorax and abdomen, yielded higher accuracy.
- InceptionV3 demonstrated the best classification performance, with Grad-CAM confirming biological relevance.
Conclusions:
- SolenopD effectively integrates YOLOv11 for precise localization and InceptionV3 for accurate Solenopsis spp. classification.
- The system supports taxonomic identification, potentially improving invasive species management.
- Segment-based analysis enhances classification accuracy, aligning with traditional taxonomic methods.
