pySilver/GLAMI-1M-remapped

GLAMI-1M数据集包含110万时尚商品，包含96.8万张唯一图片和100万段唯一文本描述。这个数据集涵盖13种语言，主要是欧洲语言，并包含191个细粒度类别，例如15种不同类型的鞋子。数据集提供了高质量的注释，由专业策展人提供，同时呈现出一个困难的产业问题。每个样本都包含一个图像、国家代码、相应语言的名称、描述、目标类别和标签来源，标签来源可能是人工或基于规则的，但大多数样本是基于人工的。

GLAMI-1M dataset contains 1.1 million fashion items, including 968,000 unique images and 1 million unique text descriptions. The dataset covers 13 languages, mostly European, and is divided into 191 fine-grained categories, such as 15 different types of shoes. It provides high-quality annotations from professional curators and presents a challenging industrial problem. Each sample includes an image, country code, name in the corresponding language, description, target category, and label source, which can be either human-based or rule-based, but most samples are human-labeled.