Development of a Deep Learning Model for the Classification of Mosquito Larvae Images

  • Ramon Mayor Martins UFSC
  • Bruno Manarin Espíndola UFSC
  • Pedro Philippi Araujo UFSC
  • Christiane Gresse von Wangenheim UFSC
  • Carlos José de Carvalho Pinto UFSC
  • Gisele Caminha Laboratory Reference Network


Dengue is a disease that is endemic to certain regions, and in 2022, it was responsible for more than three million cases in the Americas. One of the most effective ways to prevent Dengue is by preventing the formation of breeding sites for the Aedes aegypti mosquito, which is the primary vector of the disease. Unfortunately, identifying these breeding sites remains a challenge as citizens lack knowledge to distinguish Ae. Aegypti larvae from other species. A solution can be the development of a Deep Learning model, to be deployed in a mobile application that classifies mosquito species using photos of larvae. Currently only a few models are available that mostly differentiate only between genera (Aedes versus non-Aedes), or present very low accuracy. Therefore, the objective of this research is to develop an image classification model that can differentiate between Ae. Aegypti, Aedes albopictus, and Culex sp. Larvae using pictures taken with a cellphone camera by comparing various Deep Learning models (Mobilenetv2, ResNet18, ResNet34, EfficientNet_B0 and EfficientNet_Lite0). Best results were obtained with EfficientNet_Lite0 with an accuracy of 97.5% during validation and 90% during testing, an acceptable result considering the risks related to a misclassification in this context. These results demonstrate the viability of a classification of mosquito larvae differentiating even between Aedes species and thus providing a contribution to the prevention of dengue.

MARTINS, Ramon Mayor; ESPÍNDOLA, Bruno Manarin; ARAUJO, Pedro Philippi; WANGENHEIM, Christiane Gresse von; PINTO, Carlos José de Carvalho; CAMINHA, Gisele. Development of a Deep Learning Model for the Classification of Mosquito Larvae Images. In: BRAZILIAN CONFERENCE ON INTELLIGENT SYSTEMS (BRACIS), 12. , 2023, Belo Horizonte/MG. Anais [...]. Porto Alegre: Sociedade Brasileira de Computação, 2023 . p. 129-145. ISSN 2643-6264.