Acerca de Técnicas de Aumento de Dados para a Detecção Automática de Focos de Mosquito

Abstract

This work discusses data augmentation techniques for detecting mosquito breeding grounds using videos recorded by a drone. A computer-vision system is proposed for automatically detecting mosquito-breeding related objects. A database composed of six aerial videos containing breeding-related objects, such as water tanks, tires, and bottles, is devised. Due to the difficulty of obtaining extensive records of real scenarios, artificial data augmentation techniques are presented, and three methods for inserting objects into videos are considered. A convolutional neural network detector is used to evaluate these techniques, indicating that artificial data augmentation reduces overfitting, improving the overall detection performance. Keywords— data augmentation, object detection, mosquito breeding grounds