Photo Cultures

Photography is not a single phenomenon, and aesthetic perception is not universal.

The Photo Cultures projects uses methods from visual arts, computer vision, and machine learning to compare visual activity of Instagram images at different locations in 3 continents.

Culture: characterizes behaviors, beliefs or artifacts of a group of individuals in a particular time period and location(s).


Neues Sehen.jpgPhoto Culture. Artistic movement within the  “photography” phenomenon,  with its own set of distinct aesthetic rules and defining mechanisms. Examples of photo cultures include the “New Vision” European photography in late 20s, the socially conscious photography practiced in New York in the 30s, or the snapshot-style fashion photography popular in the 90s.



Can we find photo cultures in online photo sharing platforms such as Instagram?

Thanks to the diversity of the technology used, we discover the essence of photo cultures and present findings about their typical stylistic attributes, stereotypical subjects and aesthetic uniqueness of online photo cultures around the world. A new visualization method specifically tailored for photo culture studies allows qualitative studies of aesthetics around the world. And technical essays explain the scientific work behind all this.

Panel 1


Data Collection

A dataset of images shared from all around the world.

To find “photo cultures” and  explore their similarities and differences at
different locations, we selected five cities located on 3 continents:

Bangkok, Berlin,Moscow, Sao Paulo, and Tokyo.

We carefully selected cities that are very different in material and objective ways. We use the Instagram API to collect images and their metadata in for each city, over a single full week: Dec 4–Dec 12, 2013.  We took a random sample of 20K images from each city.

Data Description through Computer Vision

A dataset of images shared from all around the world.

To understand photo cultures, we characterise various aspects of each image:


SUBJECTS. Through computer vision techniques, we extract a group of Subject features that describe the image objects and scenes. We manually organize object-level tag names into a smaller set of 14 groups, according to their semantics: architecture, artifacts, fashion, furniture, tools, vehicles, animal, natural, plants, humans, food, activities, concepts, other.
STYLES. Based on computational aesthetics detectors, we extract stylistic features that describe the image photo techniques and styles, such as color distribution, sharpness, etc.


Panel 2


To  visual exploration of photo cultures, we design a novel visualization techniques specifically tailored for culture analytics studies.

Our technique is designed to reduce dimensionality while preserving location, aesthetic pattern, and subject distribution information. It works as follows.

  1. Preserving image subject: we partition images according to the subject depicted, leading to 14 different groups.
  2. Preserving image style: we cluster images for each subject according to the stylistic features using hierarchical clustering. This allows us to group images according to the dominant stylistic choices that photographers make when representing their subjects.
  3. Grow Entourage plots. We cluster centroids in the 95-d feature space, and then compute, for each image in a cluster, its Euclidean distance from the centroid. We then project the cluster centroids to 2D using t-distributed stochastic neighbour embedding (t-SNE), giving each cluster (“entourage”) a location on a 2D map
  4. Preserving image location: we add a color tag according to location – Bangkok is green, Berlin is blue, Moscow red, Sao Paulo purple, and Tokyo yellow.


Our full set of 14 high-resolution visualizations showing all subject groups and style clusters can be found here .

Panel 3


Unsupervised Discovery

Based on the clustering-based visualization technique, we discover early signs of photo cultures existing in images shared around the world.

Subject uniqueness.

We partition images according to the subject depicted, ending up with 14 image subsets of food, people, etc.. By counting the proportion of images depicting a given subject for each city, we  start observing that localized photo cultures actually exist!

Tokyo’s users tend to take pictures of food more than users in other cities: around 50% of pictures containing food have been taken in Tokyo, 12% in Sao Paulo, 8% in Moscow. We find that subjects such as architecture are more popular in Berline, while Bangkok’s users tend to prefer fashion.

Style uniqueness.

We then use hierarchical clustering over stylistic features to group images in each subject according to their stylistic patterns.  By analysing these clusters, we can find some information about the style uniqueness of photo cultures. To do so, we look at city-specific clusters, namely stylistic clusters where at least 3/4 of the images come from the same location. Photo cultures with high number of city-specific stylistic clusters will likely be more unique in terms of style. We find that around 5% of clusters are city-specific, out of which 42 % belongs to Tokyo, 42% to Bangkok, thus suggesting photo styles in Tokyo and Bangkok are more aesthetically unique than in other cities.

Supervised Discovery

We design a multi-class classification problem able to expose pair-wise similarities and differences between photo cultures. 

Intuition. If visual patterns of a city photo culture are clearly distinguishable from others, it will be easy for a classifier to identify the correct location of the image groups drawn from that city. On the other hand, the classifier will misclassify the image groups of cities with similar visual patterns.

Experiment. From the images of a given city, we randomly sample distinct groups of 10 images. We then characterize each image group with the group average stylistic and the subject features. Next, we label each image group with a category corresponding to the city they have been sampled from, and train a multi-class 10-tree random forest classifier that predicts the location  of image groups.

Confusion matrix of the supervised discovery experiment.

Subject Features

  • Tokyo is the photo culture more easily identifiable by our classifier: it therefore has the most distinctive photo culture  in terms of subjects.
  • Moscow has the least distinguishable photo culture,
  • Bangkok and Sao Paulo are the most similar 
  • Bangkok and Berlin the most dissimilar

Stylistic Features

  • Bangkok’s photo culture has the most unique photo styles, followed by Tokyo.
  • Moscow and Berlin most similar.
  • Bangkok and Berlin the most dissimilar.
Panel 4


Stereotype Discovery

What makes photo cultures unique in terms of the subjects and styles used? What are the stereotypical subjects and styles? To address these questions, we perform a correlation-based analysis.

Methodology. We want to discover the most stereotypical visual patterns. We look at  correlations between visual features and image location. The higher this correlation between feature and city, the higher the unique presence of a given subject or style in the photo culture of the city.

To do so, we correlate each feature with 5 binary city vectors, one for each location in our dataset. For a city c, for each image I in our data, the binary city vector will be 1 if the location of image I corresponds to c, and 0 otherwise

final_subj_horiz-1Stereotypical Subjects

  • BANGKOK: clothing, dress
  • BERLIN: building, house, palace
  • SAO PAULO: face, friends, people.
  • MOSCOW: night, snow
  • TOKYO: food, meal, meat, soup


Stereotypical Styles

  • TOKYO: saturation, colorfulness
  • BERLIN: monochrome
  • BANGKOK: bright, unique, unbalanced exposure



Panel 5



Contact us!