The first step in this project was to locate single selfie photos. To do this, Selfiecity teamed up with Gnip, a software company that helped them collect and access Instagram photos. After gathering 656K images from all the cities, they then selected 1,000 from each city. To verify that these were single images, they submitted the images to Mechanical Turk workers to confirm that each photo was a single selfie, but also to tag the photos and guess the age of the individuals in them. Lastly, one member of the project team went back and examined all the photos manually. In the end, there were 650 photos from each city.
One critical step in this process was the manual examination of the selfies by someone from the project team. This step was critical because, after using Mechanical Turk workers, most of the photos were tagged correctly, but there were some mistakes. Manually reviewing the photos helped the researchers catch mistakes, which allowed them to use more accurate photos, with the final set containing 650 photos from each city. While the website project itself touches on its methods, it does not go in depth. The publication gave more background on how the data was collected from start to finish.