Detecção e remoção automática de objetos indesejados em imagens utilizando aprendizagem profunda

Abstract

Technological advances have been growing over the decades and technology is becoming more and more present in our daily lives. With this, the ways of generating and acquiring information are increasingly diverse and more accessible, generating a massive amount of data. This massive amount of data makes it possible to extract information from various techniques, for the most diverse purposes. With this in mind, this work presents an application that uses Computer Vision techniques together with Deep Learning techniques as a way to protect the privacy of one of these types of data: images. Performing the removal of unwanted objects, without having a significant impact on the observer. Suck tool that can be used for leisure, in the correction of images and removal of objects, being also an alternative alternative to tools already existing as in PHOTOSHOP®. The system showed that, compared to the tools present in PHOTOSHOP®, it manages to obtain good results to remove objects in images, with minimal user participation, requiring a much lower technical level and generating the result in a very short time.