Content:
Images: 6000 images sourced from the public domain.
Diversity: The dataset includes individuals of various ethnicities, ages, and regions to ensure comprehensive coverage.
Classes: There are 20 different classes of accessories, including masks, scarves, and face shields. Additionally, it includes classifications of faces:
With a mask
Without a mask
With an incorrectly worn mask
The full class list is available in the meta.json file.
Acknowledgements:
The images were collected and annotated by the refugee workforce of Humans in the Loop, based in Bulgaria. The annotation process was facilitated using the Supervise.ly platform. This initiative not only contributes to the fight against COVID-19 but also provides meaningful employment and skill development opportunities to refugees, empowering them through valuable work experience.