Bias-Aware Face Mask Detection (BAFMD) dataset

Bias-Aware Face Mask Detection (BAFMD)数据集是由伊斯坦布尔技术大学和卡塔尔计算研究机构共同创建，旨在解决现有面部口罩检测数据集中的种族、性别和年龄偏见问题。该数据集包含来自全球Twitter用户的6,264张图片，涵盖了13,492个戴口罩的面部和3,118个未戴口罩的面部，特别关注了代表性不足的种族和年龄群体。数据集的创建过程涉及使用先进的面部检测器筛选图像，并手动标注面部和口罩使用情况。BAFMD数据集的应用领域主要集中在提高面部口罩检测系统的准确性和公平性，特别是在全球大流行病如COVID-19期间，帮助监控公共区域的口罩使用情况。The Bias-Aware Face Mask Detection (BAFMD) dataset was co-created by Istanbul Technical University and Qatar Computing Research Institute, aiming to address the racial, gender, and age bias issues in existing face mask detection datasets. This dataset contains 6,264 images sourced from global Twitter users, covering 13,492 masked faces and 3,118 unmasked faces, with special focus on underrepresented racial and age groups. The dataset creation process involved using advanced face detectors to screen images and manually annotating faces and mask-wearing status. The main application scenarios of the BAFMD dataset are focused on improving the accuracy and fairness of face mask detection systems, particularly helping monitor mask usage in public areas during global pandemics such as COVID-19.

搜集汇总

背景与挑战

背景概述

BAFMD数据集是一个针对面部口罩检测的偏见感知数据集，旨在解决现有数据集中存在的种族、性别和年龄偏见问题。它包含来自全球Twitter用户的6,264张图片，覆盖了13,492个戴口罩和3,118个未戴口罩的面部，特别关注代表性不足的群体。该数据集主要用于提高口罩检测系统的准确性和公平性，支持全球大流行病期间的公共区域监控应用。