Mosquito Trax

Thrive Sri Lanka

Mosquito Trax

An AI-powered sensor network to track disease-carrying mosquitoes in rural Sri Lanka. ​​

San Jose, CA, USA

USA

Nonprofit

Mosquito borne diseases kill over 500,000 people annually. Climate change has exacerbated this issue, with weather patterns changing and causing changes to mosquito patterns and habitats. In countries like Sri Lanka, mosquito borne diseases are a notable public health issue, aggravated by the fact that 80% of the country’s population lives in rural communities, hard to reach for mosquito technicians. Even though Sri Lanka, where Malaria was painstakingly reduced to nominal levels, Dengue is still an issue. Tracking mosquitoes is a large part of any solution, but contains with it risks and costs, particularly in rural areas where mosquitoes thrive among water bodies and lush greens.

We have prototyped a low-cost, low-power mosquito detection and classification sensor system. Able to detect mosquitoes via sound and video, Mosquito Trax is a self-sustaining device that can operate for months without battery recharge, enabling data gathering in rural areas with minimal human involvement. The Mosquito Trax sensor can form the backbone of a sensor network that can enable island wide mosquito tracking, providing critical data to inform disease preparedness efforts. We have developed algorithms leveraging audio featurization and Mel Cepstral Coefficients followed by ML classification to differentiate mosquitoes from background sounds (including other insects) and classify mosquito types (such as Aedes that is a carrier for Dengue). Our video/ image research has developed methods for using YOLOv8 to distinguish mosquitoes from background objects, classify mosquitoes based on the identified bounding box images, and detect mosquitoes based on frame-by-frame movement patterns in videos. We have also developed low power variants of these algorithms and a multi-level prediction architecture that conserves battery power via intelligent use of camera time. Our research has been published in IEEE Conference on Big Data Analytics 2024, SPIE Workshop on Applications of Digital Image Processing 2024, and presented at Women in Machine Learning NeurIPS 2024.

As of 2023, approximately 80.79% of Sri Lanka's population—around 17.8 million people—reside in rural areas, reflecting the country's strong agricultural heritage and the continued significance of rural livelihoods. This high rural population percentage underscores the importance of addressing public health concerns that disproportionately affect these communities (see Trading Economics at https://tradingeconomics.com/sri-lanka/rural-population-percent-of-total-population-wb-data.htm). Dengue fever is a major mosquito-borne disease impacting Sri Lanka, with the country experiencing its largest outbreak in 2017, recording 186,101 suspected cases and 440 deaths. The disease has become endemic across all districts, with significant outbreaks occurring every three to four years. In 2023, the incidence rate was 407.5 cases per 100,000 people, and while the annual case fatality rate remains below 0.1%, the disease poses a persistent public health challenge. Efforts to combat dengue include initiatives like the World Mosquito Program's release of Wolbachia-infected mosquitoes in areas such as Colombo and Nugegoda, aiming to reduce transmission rates and protect vulnerable populations. Our solution will assist Sri Lanka’s mosquito control efforts by enabling low-cost monitoring in rural areas. Edge devices can be deployed to gather mosquito data without human intervention, reducing both infection exposure risk and the cost and complexity of manual monitoring and intervention, particularly in hard-to-reach communities.