Are you an Instagrammer? It's highly likely that you are, happy to share your life with the world filtered through the sepia tones of bygone eras. But as you snap away, have you ever thought about the collective vision you might be adding to? The big data picture? Like, say, the visual characteristics of your city, as envisioned by the collective mass of all the Instragram photos taken there?
Chances are the answer to those questions is: no. Because if you you were thinking that, you'd probably be part of new media maven Lev Manovich, Jay Chow, and Ph.D. student Nadav Hochman's team who have visualized 2.3 million Instagram photos from 13 cities across the globe in their Phototrails study. The results of which are published in journal First Monday.
Visual signatures of San Francisco (left) compared with Bangkok (right). Based on 50,000 Instagram photos organized by brightness median (radius) and hue median (perimeter).
There idea of the study was to see what these pics might say about cultural and social patterns around the world. "The goal is not to solve 'practical' problems" Manovich says, "but to explore this new visual universe of 'social photography.'" So what sort of patterns did they find? One that they noticed quite quickly is how each city has a unique “visual signature” which they based on hue, brightness, line orientation and other factors. They also noticed the habits of users, the frequency with which they took photos and how this would change on special occasions, like a festival in the city.
As well as these characteristics, other noticeable patterns are when major events happen, like Hurricane Sandy in New York. In the study they downloaded all the photos taken in Brooklyn during the day of the hurricane. "This visualization reveals the intensity of the event," they note "manifested in the sudden dramatic decrease in the number of photos and their different colors after the power went out. We see a clear line that separates 'before' and 'after' the event."
But it's interesting to see what other conclusions can be drawn with this "big visual data". "While in the last few years we see many projects that visualize text patterns in social media (for instance, the languages used on Twitter in the NYC area) or metadata (mapping locations of tweets in NYC). Surprisingly nobody looked at larger image or video samples from media sharing sites." Manovich says.
Montage visualization of 57,983 photos from NYC taken between the 18th and the 25th of February 2012. Photos are sorted by upload date and time (top to bottom, left to right).
So, does that mean an Instagram photo is worth a thousand tweets? Manovich and Hochman believe the content of the images can be very revealing—like whether the photo contains people or faces, the angle of the camera, composition, and colors. "All this tells us about the person taking a photo." explains Manovich. "And we can compare these attributes across lots of photos from different places and different users and see how they correlate to demographic information, where people live (size of a city, country, etc.) and so on."
Close up of radial image plot visualization of 11,758 photos shared on Instagram in Tel Aviv during April 25-26, 2012. The photos are sorted by hue (radius) and time created (angle).
As with previous visualizations Manovich has been involved with, these aren't plotted as graphs or points, instead they show the actual images. The pair believe this enables analysts to see patterns you would otherwise miss. At the moment some of the patterns revealed seem whimsical, like the fact that New York is bottom when it comes to using filters (which you'd think was a major point of using Instagram), while Tel Aviv, London, and San Francisco have the highest percentage of users who use filters. What can this tell us? Well, maybe that New Yorkers are too busy to use filters. Or maybe they've just become jaded with everything looking like it was shot 40 years ago.
Radial image plot visualization of 33,292 photos shared on Instagram in Tel Aviv during 20–26 April 2012. The photos are sorted by hue (radius) and upload time (angle).
But the deep analysis of these visualizations hasn't really begun yet. Looking to the future Manovich says you could look at how factors—neighborhoods, lifestyles, and populations—change over time, along with looking at differences in different parts of a city that aren't based on standard demographic and economic statistics. Another aspect is the data is much "warmer" compared to the cold statistics of demographic and economic data. The photos are subjective, personal, so can maybe give us a deeper understanding, when used in conjunction with other data, of the people who are using these applications and the cities they inhabit. Another use, Manovich says, could be creating new types of city maps, "Representing patterns of people's experiences, feelings, and imagination, rather than zip codes and streets." Something similar to what Moritz Stefaner has already begun exploring.
This is just the beginning for the project, the team also want to look for patterns using algorithms, to look at the content of social media photos classifying them depending on what the image shows: photos of food, cities, outdoor, etc. And also to look at a larger number of photos. "Like one billion." says Manovich. "We are currently exploring how we can do that."
Images courtesy of Phototrails