Visão computacional para detecção e segmentação de resíduos sólidos urbanos

Abstract

Proper management of urban waste is a global challenge due to the growing volume of waste generated daily and the negative impact this has on human health and the environment. According to the UN, most products we consume are discarded quickly, resulting in an alarming and disorganized amount of waste. In addition, the growing demand for resources and improper waste disposal contribute to environmental degradation. This study proposes using Computer Vision techniques to detect and segment different types of waste in urban environments. The proposed Deep Learning model produced results compatible with state-of-the-art and surpassed the performance of other methods analyzed regarding a collection of images representing urban scenarios, mAP 65,9%. On the other hand, the proposed model could not properly handle small objects and non-urban scenarios.