Find Out If Your Photo Is in This AI Training Dataset
Facial recognition systems are everywhere, from security cameras trying to detect criminals to the way Snapchat finds your face to wear bunny ears. Computers need a lot of data to learn how to recognize faces, and some of it comes from Flickr.
IBM released the Face Variety dataset earlier this year, which is probably a good thing in some ways: Many early facial recognition algorithms have been trained on skinny white celebrities because it’s easy to find a lot of celebrity photos. … Your data source affects what your algorithm can do and understand, which is why there are many racist, sexist algorithms out there . This dataset is intended to help by providing facial images along with facial data such as skin color.
But most of the people who uploaded their personal snapshots to Flickr probably didn’t realize that years later, their faces, and those of their friends and families, could be used to teach the next big mega-algorithm. If you have applied a Creative Commons license to your photographs, even “non-commercial” ones, you might be in this dataset.
NBC says IBM says it will remove images from the dataset at the request of the photographer or the person being photographed, but they haven’t made the dataset publicly available, so there’s no way to know for sure if you’re actually there. Removing a photo won’t be easy, but if you want to know if any of your photos have been used, you can enter your Flickr username into the NBC tool here . This is not necessarily the only data your photo may contain, but at least there is a way to tell if your photos have been used.