WineSensed
收藏资源简介:
WineSensed是一个大型多模态葡萄酒数据集,用于研究视觉感知、语言和风味之间的关系。该数据集包含897k张葡萄酒标签图像和824k条来自Vivino平台的葡萄酒评论。它涵盖了超过350k种独特的葡萄酒,注释了年份、地区、评分、酒精百分比、价格和葡萄成分。通过进行葡萄酒品尝实验,对一部分数据进行了精细的风味注释,参与实验的256名参与者根据风味相似性对葡萄酒进行排名,产生了超过5k对风味距离。
WineSensed is a large-scale multimodal wine dataset designed to investigate the relationships between visual perception, language, and flavor. This dataset contains 897,000 wine label images and 824,000 wine reviews sourced from the Vivino platform. It covers over 350,000 unique wines, annotated with vintage, region, rating, alcohol percentage, price, and grape composition. Fine-grained flavor annotations were performed on a subset of this dataset via wine tasting experiments: 256 participants ranked wines based on flavor similarity, generating over 5,000 flavor distance pairs.




