Revisão bibliográfica sobre a capacidade de heteroidentificação dos algoritmos de biometria facial

Abstract

This work aims to identify flaws in the ability of heteroidentification in facial recognition algorithms, through systematic review of works that have already reported these errors. Pointing out how to collect data for the repositories used in this technology and how data balancing is important to stop the errors obtained in this development. It will be used articles, books, blogs, research sites, documentaries to identify the importance of risk management in this process, concepts of facial recognition, racism. Analyzing how failures in this technology can affect black individuals, since they are not only the most physically violated but also suffer online through flaws in the algorithms due to insufficient data trained by artificial intelligence. Since they need concise, coherent, and sufficient data to successfully perform what they have been taught, thus being able to give positive results and covering the whole, not just the majority