DiveFace

DiveFace包含在与性别和种族 (男性，女性和三个种族) 相关的六个类别中均匀分布的注释。性别和种族已按照半自动程序进行了注释。有24k身份 (4k类)。每个标识的平均图像数量是5.5的，对于大于150K的总图像，最少3个。用户根据其性别 (男性或女性) 和与种族物理特征相关的三个类别进行分组: 东亚人 (第1组): 祖籍在日本、中国、韩国等地区国家的人士。 撒哈拉以南和印度南部 (第2组): 在撒哈拉以南非洲，印度，孟加拉国，不丹等地区具有血统的人。 高加索人 (第3组): 欧洲，北美和拉丁美洲血统的人 (欧洲血统)。DiveFace contains annotated samples uniformly distributed across six categories tied to gender and race (male, female, and three racial groups). Annotations for gender and race were performed using a semi-automatic procedure. The dataset comprises 24k unique identities (4k classes), with an average of 5.5 images per identity and a minimum of 3 images per identity, yielding a total of over 150k images overall. Participants are categorized based on their gender (male or female) and three racial groups defined by physical characteristics: 1. East Asian Group (Group 1): Individuals with ancestral origins in countries and regions including Japan, China, South Korea, and similar areas. 2. Sub-Saharan African and South Indian Group (Group 2): Individuals with ancestry in regions such as sub-Saharan Africa, India, Bangladesh, Bhutan, and other comparable regions. 3. Caucasian Group (Group 3): Individuals with ancestry from Europe, North America, and Latin America (of European descent).

搜集汇总

背景与挑战

背景概述

DiveFace是一个包含性别和种族注释的人脸数据集，涵盖六个类别，总图像超过150K，涉及24k个身份。性别分为男性或女性，种族基于物理特征分为东亚人、撒哈拉以南和印度南部、高加索人三个组别，注释通过半自动程序完成。