Computer Vision Driven Assistive Solution for the Visually Impaired/Blind Users

Abstract

In recent years, has been observed an increase of people with a problem with one of the fundamental parts of the human body: the eyes. This thesis aims to analyze and test state-of-the-art lightweight object detectors in embedded system as Raspberry PI4 and use them to help people with the mentioned issues. The thesis has been gone further with the training using transfer learning and testing different configurations with frozen models in different use cases.