Modeling Thousands of Human Annotators for Generalizable Text-to-Image Person Re-identification | Read Paper on Bytez