Medical mask dataset
More than 6000 images for detecting masks and accessories
Humans in the Loop is publishing an open access dataset annotated as a contribution to the worldwide fight against COVID-19.
The dataset consists of 6k images acquired from the public domain with an extreme attention to diversity, featuring people of all ethnicities, ages, and regions.
In addition, the datset covers 20 classes of different accessories as well as a classification of faces with a mask, without a mask, or with an incorrectly worn mask.
The images were collected and annotated by the refugee workforce of Humans in the Loop in Bulgaria.
This medical mask dataset is dedicated to the public domain by Humans in the Loop under CC0 1.0 license 
Dataset size
A total of 6024 images.
Classes
The images are labeled in 20 classes, including:
1. face_with_mask
2. face_with_mask_incorrect
3. face_no_mask
4. face_other_covering
And 16 other classes for masks, face shields, hats, scarves, and other accessories.