Note: This dataset links to images on Instagram. We do not store or own the images on Instagram.
Image dataset from Instagram of people wearing medical masks, non-medical (DIY) masks, or no mask. Created using the Universal Data Tool for helping people come up with creative solutions for COVID-19 problems. The dataset currently has roughly ~1,205 image samples.
This dataset could be used to build a face mask detector for selfie-type photos.
Click this to download mask_labels.udt.csv. (you might have to "Save Page As" to save it)
It looks kinda like this:
| path | . | imageUrl | output |
|---|---|---|---|
| samples.0 | . | https://example.com/image1.jpg | medical_mask |
| samples.1 | . | https://example.com/image2.jpg | no_mask |
| samples.2 | . | https://example.com/image3.jpg | non_medical_mask |
Download the mask_labels.udt.csv file and open it at universaldatatool.com
Check out this notebook to see how to use the dataset with fast.ai. You could also download the mask_labels.udt.csv file to use whatever framework you want.
We were able to achieve 93% accuracy prediction accuracy on the dataset (see confusion matrix below):
