Fashionpedia-Taste

Fashionpedia-Taste数据集由康奈尔大学和哥本哈根大学创建，旨在解释人类时尚品味。该数据集包含10,000条表达（对图像的喜欢或不喜欢），52,962个选定属性，20,000个人类注意力注释和20,000个标题，覆盖1500张独特图像和100个独特主题。数据集创建过程中，通过收集用户的个人偏好信息，如性格和偏爱的时尚品牌，来丰富数据集内容。该数据集主要用于构建计算模型，以从不同的人文视角和模式中全面理解和解释消费者的时尚品味，从而解决时尚电子商务中的推荐系统问题。

The Fashionpedia-Taste dataset, created by Cornell University and the University of Copenhagen, aims to explain human fashion taste. This dataset contains 10,000 expressions (likes or dislikes towards images), 52,962 selected attributes, 20,000 human attention annotations and 20,000 captions, covering 1,500 unique images and 100 unique topics. During the dataset creation process, user personal preference information such as personality and preferred fashion brands is collected to enrich the dataset content. This dataset is primarily used to construct computational models for comprehensively understanding and interpreting consumers' fashion taste from diverse humanistic perspectives and patterns, so as to address recommendation system issues in fashion e-commerce.