传统的时尚行业面临着大数据冲击，通过大数据技术创新是提高时尚行业整体竞争力的关键步骤。运用机器学习和计算机视觉等技术对海量的时尚数据进行语义分析与理解，用以时尚流行趋势分析、支持多种信息检索方式和时尚搭配推荐，是时尚行业的大数据分析的核心技术之一，具有重大的研究意义。本项目拟针对时尚数据的特点，对时尚数据展开多个维度的表征学习，包括基于深度学习的半监督时尚图片理解算法、面向多模态多领域时尚数据的跨媒体检索算法和基于深度度量学习模型的时尚搭配算法，并最终形成一个完整的面向时尚行业的大数据分析平台。本项目的研究能帮助时尚行业相关人员有效地理解、管理、搜索和分析时尚数据，探索了时尚行业的智能化转型。

As traditional fashion industry is facing grand challenges from big data, technology innovation is a crucial step to improve the industrial competitiveness. Understanding and analyzing the semantics of large-scale fashion data through machine learning and computer vision techniques is one of the essential business analytics and technology tools, which is significantly important to support fashion trend analysis, fashion information retrieval and fashion recommendation. Based on the characteristics of fashion data, this project aims to study multiple representation learning approaches for fashion data, including deep learning based semi-supervised fashion image understanding algorithm, cross-media retrieval algorithm for multi-modal and multi-domain fashion data and deep metric learning based fashion outfit composition algorithm. Eventually, it enables to develop a big data analytics platform for fashion industry. This study would help people from fashion industry to understand, manage, search and analyze fashion data very effectively, which boosts fashion industry towards the intellectualized era.