Garment-Lookbook Pairs Dataset
收藏资源简介:
本文介绍了一个名为“Garment-Lookbook Pairs Dataset”的大规模数据集,该数据集由苏黎世联邦理工学院的研究人员构建,旨在支持虚拟时尚摄影任务。数据集包含超过55万张时尚杂志和走秀图片,以及约950万张来自电子商务平台的服装图片。这些图片通过自动化检索管道进行匹配,形成了不同质量级别的服装-杂志对,包括高质量(10000对)、中等质量(50000对)和低质量(300000对)。该数据集为生成具有创意、氛围和情境感的时尚图片提供了基础,有助于设计人员和时尚品牌自动生成创意摄影材料。
This paper introduces a large-scale dataset named "Garment-Lookbook Pairs Dataset", constructed by researchers from ETH Zurich, which aims to support virtual fashion photography tasks. The dataset contains over 550,000 fashion magazine and runway images, as well as approximately 9.5 million clothing images sourced from e-commerce platforms. These images are matched via an automated retrieval pipeline to form garment-magazine pairs at different quality tiers: high-quality (10,000 pairs), medium-quality (50,000 pairs), and low-quality (300,000 pairs). This dataset provides a foundation for generating creative, atmospheric and context-aware fashion images, and can assist designers and fashion brands in automatically generating creative photographic materials.




