A video montage of 640 selfie images from London. The individual images are identically aligned with respect to eye position and sorted by the head tilt angle. The blending animation is designed to create abstractions of the individual images, but still maintains a degree of fidelity with respect to image details and context.
Investigating the style of self-portraits (selfies) in six cities across the world using a mix of theoretical, artistic and quantitative methods.
This is a special edition of the original selfiecity project for the Big Bang Data exhibition, Somerset House, London.
We collected and analyzed 152,462 Instagram images from central London for the week of September 21–27, 2015 and compared the results to our findings from five other cities. Experiment with the new data in the selfiexploratory, and scroll down for an overview of the London findings.
First, here is what we found about how London selfies compare to other cities in terms of demographics, using a mixture of automatic methods and human judgments:
Using automatic face recognition software, we also learned that compared to New York, Moscow, Berlin, Bangkok, and São Paulo, London selfies have:
Experiment with all the data we collected.
Do angry people tilt their heads more strongly? And what are the typical moods of selfie takers in London? Find out!
This project is based on a unique dataset we compiled by analysing tens of thousands of images from each city both through automatic image analysis and human judgements.
To locate selfies photos, we randomly selected 140,000 photos (20,000-30,000 photos per city) from a total of 808,000 images we collected on Instagram. 2-4 Amazon’s Mechanical Turk workers tagged each photo. For these, we asked Mechanical Turk workers the simple question "Does this photo shows a single selfie"?
We then selected top 1000 photos for each city (i.e., photos which at least 2 workers tagged as a single person selfie).
We submitted these photos to Mechanical Turk again, asking three "master workers" (i.e. more skilled workers) not only to verify that a photo shows a single selfie, but also to guess the age and gender of the person.
On the resulting set of selfie images, we ran automatic face analysis, supplying us with algorithmic estimations of eye, nose and mouth positions, the degrees of different emotional expressions, etc.
As the final step, one or two members of the project team examined all these photos manually. While most photos were tagged correctly, we found some mistakes. We wanted to keep the data size the same (to make visualizations comparable), so our final set contains 640 selfie photos for every city.
We manually categorized a set of 2000 images in four genres: single selfies, selfies showing two or more faces, single portraits, and portraits with two or more people. Here is what we found:
Besides the manual tagging described above, we also used the openCV software library to automatically detect faces in all 152,462 images we collected from central London. This allowed us, for instance, to visualize the average number of faces per location.
The resulting map reveals a very interesting pattern: locations with fewer photos have much higher proportions of faces in these photos. Perhaps people are looking for smaller, more private spots for their portrait or seflie photos or simply prefer to take these photos in interesting and unique spaces with no tourists around.
In addition to Instagram, we also used a sample of images shared in London on Twitter. This data came from the unique dataset of 265 million geolocated tweets with images for 2011-2014 provided to our lab by Twitter as part of their 2014 Twitter Data Grants. The map below shows locations of 3,691,003 tweets with images in London (shared between November 2012 and July 2014) from our dataset.
This project is on display at the Big Bang Data exhibition, Somerset House, London, 03 Dec 2015 — 28 Feb 2016. The exhibit features a portrait mode adaptation of the selfiexploratory as well as a large rear projection of the selfiecity London video montage and large prints of some of the graphics seen on this site.
Expert on digital art and culture; Professor of Computer Science, The Graduate Center, CUNY; Director, Software Studies Initiative.
Independent consultant in information visualization / Truth and Beauty Operator. M.A. in Interface Design, B.Sc. in Cognitive Science.
moritz.stefaner.euResearcher Scientist, Software Studies Initiative; Ph.D. in Computational Neuroscience, UCSD.
lab.softwarestudies.com/Data visualization and mobile interaction designer, Ph.D. in Media Informatics from the University of Munich.
do.minik.usFreelance Consultant; exploring the cultural impacts of ubiquitous access to information to create new products and services. M.A. Royal College of Art.
danielgoddemeyer.comArt historian and curator; co-curated the Latvian Pavilion at the 55th Venice Biennale in 2013; Ph.D. student, The Graduate Center, City University of New York.
Visual social media researcher, PhD student, University of Pittsburgh. Project director, phototrails.net
Researcher, Software Studies Initiative; Web and Mobile Developer at Motive Interactive.
jayjchow.comThe development of selfiecity was supported by The Graduate Center, City University of New York, California Institute for Telecommunication and Information, and The Andrew W. Mellon Foundation.
Augmented-data.comperformed face detection in Instagram images from London.
And big thanks to gnip for the support with the data collection!