Collection of machine learning training data for the facial expression recognition

摘要

to run the crowdsourcing pipeline to generate the training examples of expression classifier of machine learning, equipment, methods, and the manufacture of articles. Crowdsourcing providers, according to an instruction or a target, to actively generate an image of the facial expression. Instructions or goal is, mimic the facial expressions, or look in a certain state, or an existing facial expression recognition device may be for a "break". The images are collected and evaluated by the same or different crowdsourcing providers, it image to meet the first of quality standards is checked by an expert from. Then inspection image is, as a positive example or a negative example, be used in the expression classifier training of machine learning.BACKGROUND