This Is The Algorithm That Analyzes More of 100,000 Selfies Theoretical Purposes

It is not no surprise that a social phenomenon as the selfies is rising a lot interest in very opposite directions each other. They are, of course, all those who have found a way to show to the world in this type of photography and create your own identity, on the other hand we have to those who seek a business opportunity and finally to those who wonder, and among which myself, all the thinking behind. The reason for these photographs and a more scientific study of the issue. Beyond the frivolous that might result, is a field of knowledge of the most interesting.

There are many authors and theorists who write essays about the selfies and its consequences but There are those who want to go beyond the question and seek a technical approach to analyze the picture in search of patterns. Lev Manovich, one of the best-known theoreticians on the audiovisual sector, has created along with researchers a tool to study be photographs and go beyond.

130,000 images analyzed with Mechanical Turk

With the name of SelfieCity, we have a service charge of analyze thousands of selfies published on Instagram. The study has been limited to five cities: New York, San Paulo, Berlin, Bangkok and Moscow, and using this social network are taken photographs to identify a series of patterns. The project, explains Manovich, relies on creating tool on one side to qualify those images and at the same time get data with selected samples.

The challenge for them was in the use of these images: How to extract data from them, How to make a large drawing of the selfie as a concept but at the same time giving individual relevance to each image for blemishing values… With these values on the table, established three values that would study the algorithm: what they call imageplots, selfiexploratory, and discoveries.

Imageplots refers to all values and details that define a selfie: the humor that is in are, the age of the characters, city, gender, the inclination of the head. This first analysis confirms some clichés that we already assumed: women make more selfies than men, they lean more head than they… Something very disturbing of Selfiecity is that taking 256 samples in each city, by aligning the photos so that all have the same orientation, the sum of the selfies does not produce an image that looks like a self-portrait.

The study also serves to demystify an issue about the selfies: people do not go up as many as we think. All cities only 4% of the analyzed images are self-portraits. The amount varies according to the region but on average we find that 96% of photos are another theme. For the analysis of the photos has been used Mechanical Turk to 130,000 images where it has been necessary to study and four workers to identify and classify each image.

It is worth to check it out completely to the tool and go playing with different variables to discover images to see what patterns recur more in each type of photography. It is a very analytical vision that is accompanied by a few fairly complete theoretical analysis and which serve to understand the work that lies behind Selfiecity.