FairFace: Face Attribute Dataset for Balanced Race, Gender, and Age

Abstract

Existing public face datasets are strongly biased toward Caucasian faces, and other races (e.g., Latino) are significantly underrepresented. This can lead to inconsistent model accuracy, limit the applicability of face analytic systems to non-White race groups, and adversely affect research findings based on such skewed data. To mitigate the race bias in these datasets, we construct a novel face image dataset, containing 108,501 images, with an emphasis of balanced race composition in the dataset. We define 7 race groups: White, Black, Indian, East Asian, Southeast Asian, Middle East, and Latino. Images were collected from the YFCC-100M Flickr dataset and labeled with race, gender, and age groups. Evaluations were performed on existing face attribute datasets as well as novel image datasets to measure generalization performance. We find that the model trained from our dataset is substantially more accurate on novel datasets and the accuracy is consistent between race and gender groups.

References in corpus (8)

• Equality of Opportunity in Supervised Learning
• Learning Face Representation from Scratch
• No Classification without Representation: Assessing Geodiversity Issues in Open Data Sets for the Developing World
• Max-Margin Object Detection
• InclusiveFaceNet: Improving Face Attribute Detection with Race and Gender Diversity
• Who Makes Trends? Understanding Demographic Biases in Crowdsourced Recommendations
• Understanding the Political Ideology of Legislators from Social Media Images
• Persuasive Faces: Generating Faces in Advertisements

Cited by in corpus (26)

• Learning Transferable Visual Models From Natural Language Supervision
• Multimodal Co-learning: Challenges, Applications with Datasets, Recent Advances and Future Directions
• Disentangling Hate in Online Memes
• Enhance Multimodal Transformer With External Label And In-Domain Pretrain: Hateful Meme Challenge Winning Solution
• One Label, One Billion Faces: Usage and Consistency of Racial Categories in Computer Vision
• An Automated Pipeline for Character and Relationship Extraction from Readers' Literary Book Reviews on Goodreads.com
• Gender Classification and Bias Mitigation in Facial Images
• Data Representativeness in Accessibility Datasets: A Meta-Analysis
• Measuring Hidden Bias within Face Recognition via Racial Phenotypes
• Semi-Supervised StyleGAN for Disentanglement Learning
• Gender Slopes: Counterfactual Fairness for Computer Vision Models by Attribute Manipulation
• When Fair Classification Meets Noisy Protected Attributes
• Interpreting Face Inference Models using Hierarchical Network Dissection
• Addressing Feature Suppression in Unsupervised Visual Representations
• CASIA-Face-Africa: A Large-scale African Face Image Database
• Personalizing Pre-trained Models
• Perceptual Score: What Data Modalities Does Your Model Perceive?
• Automatic Face Understanding: Recognizing Families in Photos
• The Selectivity and Competition of the Mind's Eye in Visual Perception
• Demographic Fairness in Face Identification: The Watchlist Imbalance Effect
• Matched sample selection with GANs for mitigating attribute confounding
• Evaluation of Human and Machine Face Detection using a Novel Distinctive Human Appearance Dataset
• IIITM Face: A Database for Facial Attribute Detection in Constrained and Simulated Unconstrained Environments
• Accurately Classifying Out-Of-Distribution Data in Facial Recognition
• Fairness Properties of Face Recognition and Obfuscation Systems
• Does Face Recognition Error Echo Gender Classification Error?