Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Bias-Aware Face Mask Detection Dataset

Alperen Kantarcı, Ferda Ofli, Muhammad Imran, Hazım Kemal Ekenel

TL;DR

The paper tackles demographic bias in masked-face detection by introducing the Bias-Aware Face Mask Detection (BAFMD) dataset, collected from Twitter worldwide and balanced across race, gender, and age using FairFace predictions. It employs a YOLO-v5-based detector and compares against multiple baselines, showing that BAFMD yields improved generalization across datasets and robustness to data volatility, while maintaining competitive detection performance. The authors demonstrate that a demographically balanced dataset reduces bias and enhances cross-domain applicability, supporting fairer and more reliable mask-detection in diverse real-world settings. The work provides a public resource and empirical evidence that balanced, real-world data improves fairness and effectiveness in face mask detection, with practical implications for monitoring public health compliance.

Abstract

In December 2019, a novel coronavirus (COVID-19) spread so quickly around the world that many countries had to set mandatory face mask rules in public areas to reduce the transmission of the virus. To monitor public adherence, researchers aimed to rapidly develop efficient systems that can detect faces with masks automatically. However, the lack of representative and novel datasets proved to be the biggest challenge. Early attempts to collect face mask datasets did not account for potential race, gender, and age biases. Therefore, the resulting models show inherent biases toward specific race groups, such as Asian or Caucasian. In this work, we present a novel face mask detection dataset that contains images posted on Twitter during the pandemic from around the world. Unlike previous datasets, the proposed Bias-Aware Face Mask Detection (BAFMD) dataset contains more images from underrepresented race and age groups to mitigate the problem for the face mask detection task. We perform experiments to investigate potential biases in widely used face mask detection datasets and illustrate that the BAFMD dataset yields models with better performance and generalization ability. The dataset is publicly available at https://github.com/Alpkant/BAFMD.

Bias-Aware Face Mask Detection Dataset

TL;DR

The paper tackles demographic bias in masked-face detection by introducing the Bias-Aware Face Mask Detection (BAFMD) dataset, collected from Twitter worldwide and balanced across race, gender, and age using FairFace predictions. It employs a YOLO-v5-based detector and compares against multiple baselines, showing that BAFMD yields improved generalization across datasets and robustness to data volatility, while maintaining competitive detection performance. The authors demonstrate that a demographically balanced dataset reduces bias and enhances cross-domain applicability, supporting fairer and more reliable mask-detection in diverse real-world settings. The work provides a public resource and empirical evidence that balanced, real-world data improves fairness and effectiveness in face mask detection, with practical implications for monitoring public health compliance.

Abstract

In December 2019, a novel coronavirus (COVID-19) spread so quickly around the world that many countries had to set mandatory face mask rules in public areas to reduce the transmission of the virus. To monitor public adherence, researchers aimed to rapidly develop efficient systems that can detect faces with masks automatically. However, the lack of representative and novel datasets proved to be the biggest challenge. Early attempts to collect face mask datasets did not account for potential race, gender, and age biases. Therefore, the resulting models show inherent biases toward specific race groups, such as Asian or Caucasian. In this work, we present a novel face mask detection dataset that contains images posted on Twitter during the pandemic from around the world. Unlike previous datasets, the proposed Bias-Aware Face Mask Detection (BAFMD) dataset contains more images from underrepresented race and age groups to mitigate the problem for the face mask detection task. We perform experiments to investigate potential biases in widely used face mask detection datasets and illustrate that the BAFMD dataset yields models with better performance and generalization ability. The dataset is publicly available at https://github.com/Alpkant/BAFMD.
Paper Structure (9 sections, 3 figures, 4 tables)

This paper contains 9 sections, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Proposed Dataset
  4. Masked Face Detection
  5. Experiments
  6. Same-Dataset Experiments
  7. Cross-Dataset Experiments
  8. Robustness to Volatile Social Media Data
  9. Conclusions

Figures (3)

  • Figure 1: [Best viewed in color] Example face mask images available in the proposed dataset. Unlike simulated or the pre-pandemic datasets, various colors and textures of the face masks are present.
  • Figure 2: Example images from the Bias-Aware Face Mask Detection (BAFMD) dataset.
  • Figure 3: FairFace analysis tool pipeline has been executed over all images of the BAFMD and MAFA datasets. FairFace gender, race and age group predictions for TFMD dataset are presented in \ref{['fig:statistics_subfig1']}, \ref{['fig:statistics_subfig2']}, \ref{['fig:statistics_subfig3']}, respectively. Similarly, for MAFA dataset gender, race and age group predictions are presented in \ref{['fig:statistics_subfig4']}, \ref{['fig:statistics_subfig5']}, \ref{['fig:statistics_subfig6']}, respectively.