Related Experiment Video
Updated: Jan 8, 2026
Flying Insect Detection and Classification with Inexpensive Sensors
Published on: October 15, 2014
Augmenting community-driven vector surveillance with automated image classification: Lessons from the Artificial
Monika Falk1, Joan Garriga2, Roger Eritja2
1Department of Ecology and Complexity, Centre for Advanced Studies of Blanes (CEAB-CSIC), C/ Accés Cala Sant Francesc, 14, Girona, Blanes 17300, Spain; Department of Computer Science, Faculty of Electrical Engineering and Computer Science, VSB - Technical University of Ostrava, 17. listopadu 2172/15, Ostrava-Poruba 708 00, Czech Republic.
Mosquito Alert uses AI to identify mosquitoes from public-submitted photos, improving surveillance. The system aids early detection of disease vectors and supports public health responses.
Area of Science:
- Environmental science and entomology
- Public health and epidemiology
- Computer science and artificial intelligence
Background:
- Community-driven surveillance is crucial for monitoring mosquito populations, especially invasive species.
- Manual identification of mosquito species is time-consuming and requires specialized expertise.
- Mobile applications offer a scalable platform for citizen science data collection.
Purpose of the Study:
- To evaluate the performance of the Artificial Intelligence Mosquito Alert (AIMA) system in classifying mosquito species from citizen science data.
- To assess the impact of model updates on AIMA's accuracy and efficiency over two operational periods (2023-2024).
- To determine the system's utility as an Early Warning System (EWS) for public health interventions.
Main Methods:
- Development and integration of a machine learning image classification pipeline within the Mosquito Alert platform.
- Automated triaging of geolocated mosquito images submitted via a mobile app.
- Real-time classification of submissions, flagging critical reports for expert review and providing immediate feedback to contributors.
- Comparative analysis of model performance across different mosquito species and operational periods.
Main Results:
- AIMA demonstrated reliable classification for Aedes albopictus and Culex sp. across both model versions.
- Aedes aegypti remained challenging to identify accurately, indicating areas for further model improvement.
- The system significantly reduced the expert time required for routine mosquito identifications.
- Real-time distribution maps of invasive species were generated, providing actionable data for public health authorities.
Conclusions:
- The AIMA system enhances mosquito surveillance by automating species identification, enabling scalable and responsive public health efforts.
- AI-powered citizen science platforms like Mosquito Alert are vital for early detection and management of mosquito-borne diseases.
- Continued model development is necessary to improve the identification accuracy of challenging species like Aedes aegypti.
