Related Experiment Video
Updated: Aug 23, 2025
05:16
Flying Insect Detection and Classification with Inexpensive Sensors
Published on: October 15, 2014
25.3K
A Swin Transformer-based model for mosquito species identification
De-Zhong Zhao1,2, Xin-Kai Wang2,3, Teng Zhao2
1College of Mechanical and Electrical Engineering, Beijing University of Chemical Technology, Beijing, 100029, China.
Scientific Reports
|November 5, 2022
Summary
Deep learning models accurately identify mosquito species, crucial for controlling diseases they transmit. Swin Transformer-based models achieve over 99% accuracy, improving public health strategies.
Area of Science:
- Entomology
- Computer Science
- Public Health
Background:
- Mosquitoes transmit numerous fatal parasitic and pathogenic diseases.
- Accurate mosquito species identification is essential for effective vector control strategies.
- Current morphological and molecular identification methods have limitations.
Purpose of the Study:
- To introduce deep learning techniques for automated mosquito species identification.
- To develop and evaluate a novel deep learning model for high-accuracy mosquito classification.
Main Methods:
- Construction of a balanced, high-definition dataset of 9,900 mosquito images across 17 species.
- Comparative evaluation of Convolutional Neural Networks and Transformer models.
- Optimization and selection of the Swin Transformer architecture (Swin MSI) through iterative testing and image size adjustment.
Main Results:
- The Swin MSI model achieved 99.04% accuracy and 99.16% F1-score for mosquito species identification.
- Achieved 100% accuracy in subspecies-level identification within the Culex pipiens Complex.
- Demonstrated 96.26% accuracy in categorizing novel mosquito species.
Conclusions:
- Deep learning, specifically the Swin Transformer, offers a highly accurate and promising approach for mosquito identification.
- This technology can significantly aid in the control of mosquito-borne diseases.
- The model's identification process, while effective, utilizes morphological keys distinct from human interpretation.
