Over the past decades, human facial attributes (e.g. gender, age, and race) estimation has received large amount of attention and research due to its potential applications, and variety of methods have been proposed to address it. This article is devoted to review related works comprehensively and give references for researchers. Firstly, in accordance with whether exploiting the potential correlations between these facial attributes, the existing approaches are classified into naïve and natural groups and they are reviewed within each group. Then, in terms of incompleteness of annotated labels, considered attributes, and correlations utilization, the drawbacks of existing methods are analyzed. Finally, future research directions are provided at the end of this work.