Diverse Women Dataset: Comprehensive Image Set with Self-Reported Metadata for AI Applications

**Overview**
This dataset features images and self-reported metadata from 70 Black women, offering a comprehensive set of visual and demographic data for AI model evaluation. Each participant provides multiple images, including selfie (front, left, right), full body (front, left, right), and hand photos. In addition, the dataset includes detailed metadata such as age, ancestry, skin tone (using the Google Monk skin tone scale), body measurements, hairstyle, and more. This dataset is ideal for companies developing AI models related to facial recognition, body analysis, and inclusion initiatives.

**Use Cases**
- AI Model Evaluation for Facial and Body Recognition: Train AI models to fairly identify Black women across various angles and poses.
- Inclusive Beauty and Fashion AI: Use for developing applications in fashion, beauty, and retail that need diverse, real-world body shapes, hairstyles, and facial features.
- Skin Tone and Bias Research: Analyze skin tone data and bias using self-reported skin tone and a skin tone bias test, offering insights into cultural bias in AI models.
- Medical AI Applications: Enhance AI models for healthcare and fitness tracking, particularly for Black women, using accurate metadata such as body measurements and skin tone.

**Product Details**
The dataset includes the following images and metadata fields:

Images:

- Selfie (front, left, right)
- Full-body (front, left, right)
- Hand picture

Metadata:

- Age range
- Ancestry
- Skin color (Google Monk skin tone scale)
- Skin group
- Skin tone bias test (self-selected skin tone and identification of two other skin tones)
- Weight range
- Gender identity
- Body type/shape
- Waist size
- Hip size
- Hairstyle
- Glasses (yes/no)
- Glasses type
- Head covering
- Nails status (polish, no polish, etc.)
- Natural hair texture
- Hair color

For more details, please get in touch with [e-mail]

**Additional Insights**
This dataset offers unique insights into the diversity of Black women, including comprehensive self-reported metadata on body shape, hair, and skin tone. The skin tone bias test adds an additional layer of understanding to how individuals perceive and classify their skin tone, making it useful for applications aiming to tackle bias in AI models. This dataset can be used across a wide range of applications, from beauty and fashion to healthcare, providing inclusive, real-world data for better AI development.

搜集汇总

背景与挑战

背景概述

该数据集收录70位黑人女性的多角度照片（自拍、全身及手部）及自报告元数据（年龄、血统、肤色、体型等），专为开发包容性AI模型设计，适用于面部识别、时尚美容和医疗健康等领域。其特色包含基于Google Monk量表的肤色偏差测试，可助力减少AI模型中的文化偏见。