Col-Ovo: Smartphone-based artificial intelligence for rapid counting of Aedes mosquito eggs under field conditions

Abstract Background The global expansion of Aedes aegypti mosquitoes has intensified the transmission of dengue, Zika, and chikungunya viruses, increasing the need for scalable surveillance tools. OviCol, a low-cost ovitrap-based strategy integrating mosquito monitoring and control, has shown promise for community-based vector management. However, large-scale deployment generates thousands of oviposition substrates that require labor-intensive manual egg counting, creating a major operational bottleneck. This study aimed to develop and evaluate Col-Ovo, an artificial intelligence-based system for automated counting of Ae. aegypti eggs from smartphone images collected under routine field conditions. Methods Col-Ovo was developed using a YOLOv11m object-detection model trained on 275 oviposition substrates (20.5 cm strips) collected through operational surveillance activities. Images were captured in situ using smartphones without preprocessing and included samples stained with bioattractants (blackstrap molasses and Saccharomyces cerevisiae ) as well as sand and particulate debris. The system was designed to tolerate image compression generated by messaging platforms, such as WhatsApp. Performance was assessed by comparing automated egg counts with expert manual counts and virtual-human counts conducted through the OviLab platform using >200% image magnification. Agreement between methods was evaluated using correlation and concordance analyses. Results Col-Ovo successfully detected and quantified Ae. aegypti eggs under challenging field conditions, including heavily stained and debris-contaminated substrates. Col-Ovo showed strong agreement with human evaluators (Spearman ρ = 0.91; Lin’s ρc = 0.89; MAE = 22.1 eggs) and with OviLab counts, with no significant difference in paired comparisons ( P = 0.252). The system reduced processing time from approximately 15 min per substrate to less than 3 s, representing a substantial increase in operational efficiency. Performance remained robust when analyzing compressed smartphone images shared through common messaging applications. Conclusions Col-Ovo provides a rapid, accurate, and scalable approach for quantifying Ae. aegypti eggs from smartphone images. By eliminating the need for specialized equipment or image preprocessing, the system addresses a critical barrier to large-scale ovitrap surveillance. Integrated with the OviLab platform, Col-Ovo offers a flexible digital infrastructure that can strengthen vector surveillance, support timely public health responses, and facilitate the expansion of community-based mosquito control programs.