Colors of Instagram: Can we understand the use of public spaces through the predominant colors in social networks?

Made by Paul Moscoso

Through coding and analysis tools, this project is intended to discover a clue in the relationship between green spaces in a neighborhood with the colors that predominate in Instagram images.

Created: October 29th, 2017

0

For this module, there were several social networks datasets given to interact and work. After studying the almost limitless possible outcomes of the information, I was motivated by the idea of exploring the ability of social networks to visualize urban spaces to a broader audience and if possible to assemble this information in order to understand the connection between users, networks and the physical spaces. 

0

With an Instagram database, it is possible to code in Carto SQL to stack the information of post so that it becomes comparative concerning different places, but it is limiting the fact that this social network does not georeference the content. The above image shows the different post of Instagram in different locations within Pittsburgh. 

0

So to achieve a higher degree of specificity, the project will focus on information collected in the Garfield neighborhood on the East Side of the city.

0

In order to condense the Instagram CSV file, it was decided to study green spaces and by that reduce the data to information that have a value of 1 in the Color Palette number 4, since it is where it predominates the green colors.   

0

But by reducing the list to green spaces, it arises a need to understand what are these spaces where people react and publish information within Garfield. Thus, a Foursquare database is used, and the dataset list is reduced to specific green spaces of the neighborhood, which are not redundant or not very specified.

0
SQL

WITH 
  -- group points with identical point geometries
  -- sort groups by their y-axis location so they will be rendered top to bottom
  m AS (
    SELECT array_agg(cartodb_id) id_list, the_geom_webmercator, ST_Y(the_geom_webmercator) y 
    FROM paulmoscosoriofrio.garfield_color4_dataset
    GROUP BY the_geom_webmercator 
    ORDER BY y DESC
  ),
  -- turn the groups of points back into individual rows with a number 
  f AS (
    SELECT  generate_series(1, array_length(id_list,1)) p, unnest(id_list) cartodb_id, the_geom_webmercator 
    FROM m
  )
 -- replace geometries join with original dataset on cartodb_id to get the attributes
SELECT  ST_Translate(f.the_geom_webmercator,0,f.p*12) the_geom_webmercator, f.cartodb_id, q.dominant_color_hex, q.hash_tags, q.liked_by, q.display_url, q.location_name, q.taken_at, q.caption, q.color_str
FROM f, paulmoscosoriofrio.garfield_color4_dataset q 
WHERE f.cartodb_id = q.cartodb_id
Click to Expand
0
Carto CSS

layer {
  marker-width: 7;
  marker-fill: [dominant_color_hex];
  marker-fill-opacity: 0.9;
  marker-line-color: #FFFFFF;
  marker-line-width: 1;
  marker-line-opacity: 1;
  marker-placement: point;
  marker-type: ellipse;
  marker-allow-overlap: true;
}
/** category visualization */
#stack_dummy {
   marker-fill-opacity: 0.9;
   marker-line-color: #424141;
   marker-line-width: 1;
   marker-line-opacity: 1;
   marker-placement: point;
   marker-type: ellipse;
   marker-width: 40;
   marker-height: 32;
   marker-allow-overlap: true;
}
//#stack_dummy {
//   marker-fill: [dominant_color_hex];
//}
Click to Expand
0

Using as reference the Stacking Chip approach, a script for stacking the new formatted dataset is written. To change the appearance of the information, there is also a new Carto CSS script. Check more info about Stacking Chip by Chris Whong here:  https://carto.com/blog/stacking-chips-a-map-hack/

0

In order to obtain appropriate data for the study, the foursquare information is extracted to obtain a map with the most popular green spaces.

0
0

After the systematization of the datasets, there is a final map that shows the images from Instagram picturing them by the relevant color of each image, and the location of the green spaces in Garfield. 

0

Finally, the study has determined that there is a correlation between the green spaces and the images that are placed on Instagram, at least three examples have been found that have a correlation between the colors, the check-in space, and the physical public space. The biggest limitation is that in order to use this information more effectively, it should continue to be coding in order to combine the data into a single presentation, and not empirical as it is now. It would also be interesting to propose a potential computational design that allows to automatically analyze the proposed relationships. 

x
Share this Project


About

Through coding and analysis tools, this project is intended to discover a clue in the relationship between green spaces in a neighborhood with the colors that predominate in Instagram images.