For example, one of the projects in this study was asked to classify different faces, the labelers were instructed to draw bounding boxes around the different faces, and then gave the following categories: male /Female; Age (Baby 0-2, Boy or Girl 2-16, Man or Woman 16-65, Senior 65+); Race: Caucasian, Japanese, Korean, Chinese people, Latinos, etc. Imagine that you are the person in charge of tagging. When you see a photo, what clues do you use to determine
a person's age and ethnicity? Will your upbringing, popular database cultural background, and values bring you the same judgment as another marker? Pay attention to the value of human labor in the process of data labeling These marked data are the basis of all artificial intelligence applications. When these markers are used to help speed up the identification of vehicles passing through toll booths, or to help blind people identify obstacles around them,
they are applications that improve the quality of life for everyone. But if these intelligent recognition technologies are used in monitoring systems to identify specific people, identify the emotional state of job applicants, identify a person's identity, etc., it is very important to train the machine to determine how the markers (standard answers) of the rules are generated.