Toward Fashion Intelligence in the Big Data Era: State-of-the-Art and Future Prospects

Abstract

Fashion, as a popular aesthetic expression, is often conveyed in a specific context, in particular in clothing, footwear, and accessories. Attributed to its enormous economic potential, intelligent fashion analysis has attracted focused attention from both academia and industrial practitioners in recent years. In this research, we provide a comprehensive review of fashion analysis-related tasks, which include fashion detection, fashion parsing, fashion retrieval, fashion style learning, fashion compatibility learning, fashion attribute prediction, and fashion generation. We investigated state-of-the-art articles from 1990 to the present, and provided a new taxonomy of extant research topics over these articles. We then summarized most representative fashion datasets with detailed statistical information. The coupled topics, methods, and features with respect to different levels of fashion analysis are also highlighted. Finally, remaining challenges and open issues are discussed in order to provide guidance for future studies.