Identificação de potenciais focos de dengue usando redes neurais artificiais com dados de sensoriamento remoto

This work aims to verify the feasibility of using artificial neural network (RNA) in remote sensing images to identify large dengue mosquito outbreaks, providing an automatic and effective way of combating the disease. The focus chosen for the work was the swimming pool, which has a large area, is not filled only during the rainy season and, when untreated, becomes a major breeding ground for the Aedes aegypti and Aedes albopictus mosquitoes that transmit the disease. To do this, it was necessary to define the appropriate type of artificial neural network to perform automatic detection of swimming pools using remote sensing images. The Convolutional Neural Networks (CNN) YOLOv8 network was used, due to its notoriety in image detection performance. The database for training the network was then prepared by separating the training and validation data, transforming the COCO annotations into YAML annotations, which can be correctly identified by the model. The results analyzed the performance and effectiveness of the answers obtained in the training and validation stages and observed the model’s behavior on a new dataset.