MS-Celeb-1M

MS-Celeb-1M是由微软研究院创建的大型人脸识别数据集，旨在通过网络收集的名人面部图像来识别一百万个名人。数据集包含约1000万张图像，覆盖超过200个国家和多种职业，年龄和性别分布广泛。创建过程中，数据集利用了Freebase知识图谱的唯一实体键进行名人身份的明确识别和去重。该数据集主要应用于图像搜索、排名、标题生成和深度图像理解等领域，旨在解决大规模人脸识别中的身份确定和去重问题。

MS-Celeb-1M is a large-scale face recognition dataset created by Microsoft Research, which aims to recognize one million celebrities using web-collected facial images of celebrities. The dataset contains approximately 10 million images, covering over 200 countries and diverse occupations, with a wide distribution of ages and genders. During its development, the dataset utilized the unique entity keys from the Freebase knowledge graph to explicitly identify celebrity identities and perform deduplication. This dataset is primarily applied in fields such as image search, ranking, caption generation and deep image understanding, with the goal of solving the problems of identity determination and deduplication in large-scale face recognition.