Decision Support System for Image Classification of Aedes aegypti

Dengue is one of the main arboviruses in Brazil, requiring continuous surveillance strategies for the Aedes aegypti vector. The identification of mosquitoes, traditionally carried out by specialists based on morphological characteristics, can be time-consuming and not very scalable in the face of the large volume of images from traps, laboratories and field actions. This work proposes a decision support system for entomological surveillance, aimed at the automated screening and identification of Aedes aegypti. The system integrates a classifier based on convolutional neural networks, using the MobileNetV2 architecture with transfer learning, with a Web application for image submission and visualization of results. The evaluation was carried out with public bases, in binary and multiclass scenarios. The results indicated promising performance, with accuracy close to 93% in the multiclass scenario and a good balance between precision and sensitivity. The system proved to be viable as an auxiliary tool for screening, without replacing specialized validation.