Awesome Open Source
Awesome Open Source

口罩遮挡人脸数据集(Real-World Masked Face Dataset,RMFD)



联系人:黄宝金,联系邮箱:[email protected]

为了进一步扩充数据集,欢迎大家将个人收集到的戴口罩图片,通过邮件的方式发送到 [email protected],我们会对收到的图片统一处理。


部分原始样本已上传本github站点,RMFD_part_1可直接下载使用,RMFD_part_2 (4个压缩文件) 和RMFD_part_3 (3个压缩文件) 需要下载全部压缩文件后,再进行解压。也可以从下方的地址下载:

已标注数据集说明如下:(区别于github中raw samples) 不同于人脸口罩识别(或检测)数据集,口罩人脸识别样本集须得包含同一人的多张戴口罩与未戴口罩的人脸图像,为此,我们建立了两种口罩人脸识别样本集。

(1) 真实口罩人脸识别数据集:从网络爬取样本,经过整理、清洗和标注后,含525人的5千张口罩人脸、9万正常人脸。

下载地址: 密码:j3aq


(2) 模拟口罩人脸识别数据集: 给公开数据集中的人脸戴上口罩,得到1万人、50万张人脸的模拟口罩人脸数据集。


下载地址: 密码:bts7


下载地址: 密码:q7cf


下载地址: 密码: jy5j


下载地址: 密码:ebd8


下载地址:链接: 密码:tbe6



链接: 提取码: acwe

image image image



image image image image image image image image

Real-World Masked Face Dataset(RMFD)

Because of the recent epidemic of COVID-19 virus around the world, people across the country wear masks and there appear a large number of masked face samples. We thus created the world's largest masked face dataset to accumulate data resources for possible intelligent management and control of similar public safety events in the future. Based on masked face dataset, corresponding masked face detection and recognition algorithms are designed to help people in and out of the community when the community is closed. In addition, the upgrade of face recognition gates, facial attendance machines, and facial security checks at train stations is adapted to the application environment of pedestrian wearing masks.

Sponsor: National Engineering Research Center for Multimedia Software (NERCMS), School of Computer Science, Wuhan University

Contact: Baojin Huang, Email: [email protected]

In order to further expand this dataset, everyone is welcome to send personally collected pictures of masks to [email protected] by email, and we will process the received pictures uniformly.

Download Datasets

Part of the original samples has been uploaded to this github website. RFMD_part_1 can be downloaded directly. RFMD_part_2 (4 compressed files) and RFMD_part_3 (3 compressed files) need to download all compressed files before decompressing them.You can also download these datasets from the link below.

Download link:

More labeled face samples are illustrated as follows: (different from raw samples in github) Different from the facial mask recognition (or detection) dataset, the masked face recognition dataset must include multiple masked and unmasked face images of the same subject. To this end, we have established two kinds of masked face recognition datasets.

(1) Real-world masked face recognition dataset: We crawled the samples from the website. After cleaning and labeling, it contains 5,000 masked faces of 525 people and 90,000 normal faces.

Download link: Password: j3aq


(2) Simulated masked face recognition datasets: We put on the masks on the faces in the public face datasets, and obtained the simulated masked face dataset of 500,000 faces of 10,000 subjects.

WebFace simulated masked face dataset: Password:bts7

LFW simulated masked face dataset: Password:q7cf

AgeDB-30 simulated masked face dataset: Password: jy5j

CFP-FP simulated masked face dataset: Password:ebd8

(3) Real-world masked face verification dataset contains 4015 face images of 426 people. The dataset is further organized into 7178 masked and non-masked sample pairs, including 3589 pairs of the same identity and 3589 pairs of different identities. Password: tbe6

Masked Face Recognition

Based on the constructed datasets, we designed and trained a face-eye-based multi-granularity masked face recognition model. The face identification accuracy on the dataset is over 95%, and some real-time video demos are as follows:

Download link: Password: acwe

image image image

Related Work

Examples of Raw Samples

image image image image image image image image

Alternative Project Comparisons
Related Awesome Lists
Top Programming Languages

Get A Weekly Email With Trending Projects For These Topics
No Spam. Unsubscribe easily at any time.
Python (863,999
Dataset (33,171
Face (11,984
Recognition (10,762
Face Recognition (2,531