: Standard models often fail if a person changes clothes. Newer methods aim to learn "clothes-invariant" features, such as body shape or gait, to recognize the person regardless of their outfit. Attribute-Guided Learning