LANDSCAPE THROUGH COLOR
DATA VISUALIZATION FROM 1 YEAR OF INSTAGRAM PICTURES
Project developed for the course "CS171: Visualization"
Harvard University, School of Engineering and Applied Sciences
Students: Alkistis Mavroeidi, Jonathan Taratuta-Titus, Steven Vasilakos
Instructor: Alexander Lex
Landscape through color is a project that analyzes Instagram pictures taken around the general area of Harvard Yard and then extracts and visualizes the primary colors. The visualization is meant as a tool of landscape exploration, as well as a way of analyzing the way people perceive their environment and how seasons or events affect this relationship.
1. GATHERING DATA
The data collection algorithm was almost entirely developed for previous research for the course "Introduction to Computational Design", Harvard University, Graduate School of Design (Fall 2014)
Project: “Pan’s Colors”: Instagram Color Analysis Application
Students: Alkistis Mavroeidi, Kritika Dhanda, Qurat-ul-ain Malick
Professor: Panagiotis Michalatos
Using C#, the script connects to the Instagram API, authenticates the developer’s id and then performs a search based on the criteria we selected and downloads the pictures it finds. We chose to download pictures based on geolocation and time, including parameters like radius and number of downloaded media. The pictures are saved in a local file, storing all the relevant information in their name and thus available for the mapping later on.
2. ANALYZING COLORS
After the pictures are downloaded, carrying the date, time, longitude and latitude information in their titles, they are imported again into the program to be analyzed. Extracting the primary color of a picture is a computationally complex problem, but we decided to use a three dimensional mapping of each pixel’s color and then average the most populated area. In more detail, the rgb values of each color are used as x, y and z coordinates in a 3 dimensional space. For every picture we analyze the color of each pixel and map it on the grid. We then divide it into 64 color bins (by creating a 4x4 division) and sort them from the most to the least populated one. By extracting the average of the color bins we get a pretty good aproximation of the primary colors of each picture in their relative order. The testing stages were first developed in the Grasshopper plugin for Rhinoceros but then the algorithm was integrated into Visual Studio, and the colors were mapped according to geolocation and filtered according to time. With small alterations, we adapted it the script so that instead of mapping the colors, the data was exported to a json file for our use in this project.
3. MAP FEATURES
(1) Geolocation mapping of color nodes
(2) Graph of 3 primary colors of visualized pictures, sorted by hue
(3) Time Slider
(4) Count per day Graph
(5) Video & Progress Book
(1) Hovering over color nodes reveals original picture with 10 primary color analysis
(2) Hovering over color visualization graphs reveals original picture with 10 color analysis
(3) Clicking on Time Slider or Count Graph sets the date accordingly
(4) 2D Brushing on Map interactively alters color visualization
4. POINTS OF INTEREST
Color Hue Hints
During the first days of November, an increase in pink and purple color hues was apparent, because of a beautiful sundown sky that was widely documented by people. Such color hue differentiations can gives us clues on events or behaviors that cannot be otherwise predicted or traced.
Some locations are particularly interesting because of the consistency of their popularity. These areas could be case studies of color change of a single theme, through the eyes of different visitors. (#JohnHarvardStatue)
Seasonal Color Range
The most profound differences in color change occur when the natural elements take over and transform the landscape significantly. This also changes the behavior of people towards their environment, as we can clearly see from the alterations in photograph thematology. In our examples, hues of red and yellow are prominent during the Fall, and trees become the real protagonists of the season to which Harvard Yard heavily bears witnesses. During Juno this past January, people were photographing the snow covered landscapes filling the color graphs with white hues, including pictures of ordinary objects and locations that were usually uninteresting to them, which we can clearly observe in the great diversion of the January color mapping
Graduation Events Clusterings
At the end of May Graduation day created some significant clustering of dark nodes in Harvard Yard, where the event was taking place. A location that is usually active, the area before Widener Library significantly increases its popularity during those days. (#CongratulationsClass2014)
Amy Poehler Visits Harvard
On January 29th, the Massachusetts-born comedian Amy Poehler parades through Mass Av to celebrate her announcement as Hasty Pudding’s 2015 Woman of the Year. The enthusiasm of people is clearly documented through the photographs and a clear path of Poehler’s movements can be traced through the Instagram pictures. #AmyPoehler