Mosquitoes and Public Health: Improving Data Validation of Citizen Science Contributions Using Computer Vision

Abstract

Citizen scientists have the potential to play an important role in combating mosquito-borne diseases such as malaria, dengue fever, Zika, and West Nile virus. While public health officials do not have the labor or resources to collect adequate spatial and temporal coverage of where, when, and what types of mosquitoes are found, these officials may be hesitant to include citizen science data because they are not confident of the data. The GLOBE Observer Mosquito Habitat Mapper (GO MHM) app was launched in 2017, and data are being collected from around the world. As part of the app, photos of mosquito larvae are submitted. Computer vision software provides a means of assisting in the validation of the images for correctness. We have developed an image recognition prototype that includes image collection, training of image classifiers, specimen recognition, and expert validation and analytics. Photos of known mosquito larvae are used to train a Deep Learning classification model. Citizen scientists submit photos, and the system conducts an analysis that indicates the probability of correct mosquito identification. New and better classification models can be trained from new data received from citizen scientists. Experts can have access to the recognition results to assess data quality and devise ways to improve the quality of GO MHM data, while public health officials can use the data to assist in the mitigation of disease outbreaks. Recommendations are made for prototype improvements.