Taking Snapshots of the Cultural Life of Cities

San Diego, Aug. 13, 2013  -- Do cities have their own visual DNA? According to new-media researcher Lev Manovich, they do, and this “visual signature” can be captured in the millions of photographs posted daily on social networks.

Prof. Lev Manovich (left) and UCSD alumnus Jay Chow in front of the large Vroom display wall in the Qualcomm Institute, displaying a visualization of 33,292 Instagram photos shared by people in Tel Aviv during one week.
Manovich leads the Phototrails project, which analyzed and compared 2.3 million photos from 13 major metropolises around the world. Each photo had been uploaded to the popular Instagram photo-sharing social network. Using digital image analysis and visualization techniques, the project attempted to derive meaning from the visual patterns, structures and dynamics.

“We concluded that each city has its own unique visual signature,” said Manovich, a computer science professor at the City University of New York, and director of the Software Studies Initiative in Calit2’s Qualcomm Institute at the University of California, San Diego. “We previously used computer tools to analyze patterns in large data sets such as one million user-generated artworks and one million Manga pages, but this is the first time that we have focused on what cultural analytics can tell us about the places where we live.”

The Phototrails study is a partnership among The Graduate Center of CUNY, the Qualcomm Institute, and the History of Art and Architecture Department at the University of Pittsburgh.

The ‘visual signatures’ of four cities (clockwise from top left): San Francisco, Tokyo, Bangkok, and New York City. Visualizations plot the photos using combinations of mean and median brightness, and mean and median hue.
New York City is one of the 13 global cities analyzed for the Phototrails project. Manovich and his colleagues collected Instagram photos over a three-month period (January-March 2012). During this period, they garnered 245,000 photoss, each of them tagged with the exact New York location and time when the photo was shared on Instagram. The largest number of photos in the data set came from San Francisco (344,000) and the smallest, Rio de Janeiro (65,000). Other cities covered by the Instagram study: Tokyo, London, Moscow, Tel Aviv, Bangkok, Sydney, Istanbul, Singapore, Paris, and Berlin.

The team’s findings were published in the July issue of First Monday*, an open-access, peer–reviewed online journal. All of the visualizations and findings are also available on the project’s Web site at http://phototrails.net.

The study integrates methods from social computing, digital humanities and software studies to analyze what Manovich calls “big visual data.” To assess differences between cities, the team used information recorded by Instagram every time a photo is shared: date and time, geographic location, and filter aplied to the photo. In addiiton, they also processed photos to extract basic visual attributes such as mean, median, standard deviation, and histograms for brightness, hue, and saturation; the number of edges; contrast; and texture measurements.

A radial image plot visualization shows 23,000 Instagram photos from Brooklyn during the day when Hurricane Sandy hit the region. The photos are organized by upload time (angle) and average hue (radius). Note the line marking a change in the number of photos and their colors, corresponding to the moment of the power outage in the area. The visualization shows a very marked demarcation in the volume of activity (decline), and hue of photos taken after the power went out.
From those statistics, the team created radial plot visualizations using 50,0000 image samples from each city. Each visualization shows all 50,000 images. In addition, Manovich and collaborators also compared the cities using all visual attributes. Bangkok was found to be the most visually different from the other cities, followed by Singapore and Tokyo.

“Our visualizations allow us to uncover the aggregated visual characteristics of each city as well as to examine the impact of exceptional events such as Hurricane Sandy,” said Phototrails investigator Nadav Hochman, a Ph.D. student in the History of Art and Architecture Department of the University of Pittsburgh.

Beyond comparing the visual signatures of 13 global cities, the study went deeper into space and time patterns to explore specific cities and key events. In the wake of Hurricane Sandy, the researchers were able to analyze 23,000 Instagram photos from the Brooklyn area shared between November 29 and 30, 2012, when Sandy hit the New York area. A radial plot visualization of the user-generated photos shows a steep drop in the number of photos posted, and changes in hue, after the power (and most lights) went out (see image at right).

Instagram photos from Tel Aviv on the eves of two national holidays (top left and right) are organized according to their locations and photo-sharing times. During the Day of Remembrance, most photos were shared in the morning and their locations cover the whole city (green dots, above left). During Independence Day, there is a burst of activity in the evening - especially in particular symbolically important areas of the city (red dots, above right). This includes fireworks captured in many photos. The image on the bottom shows these photos selected from all photos uploaded during Independence Day.
In addition to visualizing large photo samples together, Manovich and the team also compared Instagram routines of hundreds of application users. They selected users with most photos in Tel Aviv and created a separate plot for each user. The locations of photos are shown as dots, and the color of dots indicated when a photo was shared. In addition, if two photos were shared within one hour, the dots were connected by a line. The resulting visualization revealed different habits of Tel Aviv’s Instagram users. Some only upload photos in the morning, others only do it in particular parts of a city, etc.

“We were interested to learn which social, cultural or political insights might be gleaned from spatio–temporal visualizations of the Instagram photos uploaded in one city over a three-month period of time,” said Manovich. “We also zoomed into the data to examine the patterns over a week, and also within a day.

In addition to Hochman and Manovich, the Phototrails team includes a recent UC San Diego alumnus, Jay Chow, who majored in Interdisciplinary Computing and the Arts. Using the official application programming interface (API) provided by Instagram, the team “crawled” Instagram’s millions of publicly-shared photos and their metadata. “We analyzed every image and then visualized groups of images together using the software tools developed by the Software Studies Initiative in Calit2’s Qualcomm Institute,” explained Chow. “Our visualizations use different aspects of the data, and different layouts and techniques, in order to present the visual information in new and intriguing ways and to reveal patterns.”

Manovich (left) and Chow look at visual patterns created by Instagram photos on the Qualcomm Institute's Vroom display wall.
For Phototrails, Chow also developed a new version of the lab’s ImagePlot tool, which allows creation of radial visualizations. “Instead of creating very long plots where time is represented on the horizontal axis, we can now create more compact visualizations with time running along the perimeter of a circle. This new technique allowed to us compare different cities,” commented Manovich.

The study looked at global patterns of Instagram use during a three-month period. The great majority of people only uploaded one or a few photos. The proportion of users who uploaded more than 30 photos varied significantly from city to city: only 2 percent in New York versus 6.8 percent in Istanbul and 10.9 percent in Tel Aviv, the city with the highest percentage of active Instagram users.

There are also sharp differences in the extent to which users take advantage of adding Instagram’s automatic filters to their photos before uploading them. Roughly 81 percent of photos from Tel Aviv were filtered, with London and San Francisco close behind. “But New York had the lowest percentage of filtered photos – just 68 percent,” said Manovich.”Why? The pace of life is hectic in the Big Apple, so it’s possible that New Yorkers are just too busy to take the extra time to find and apply the right filter before uploading the photo to Instagram.”

*N. Hochman and L. Manovich, “Zooming into an Instagram City: Reading the local through social media,” First Monday, July 2013. 

Media Contacts

Doug Ramsey, 858-822-5825, dramsey@ucsd.edu

Related Links

Phototrails
The Graduate Center, City University of New York
University of Pittsburgh Art Department
lab.softwarestudies.comSoftware Studies Initiative
Qualcomm Institute, California Institute for Telecommunications and Information Technology
Instagram