Veröffentlicht am Schreib einen Kommentar

Treemaps and Art History?

Treemapping is a visualization method for displaying hierarchic structured data by using nested rectangles. This approach was firstly described in Ben Sheiderman’s essay Tree Visualization with Tree-Maps: 2-d Space-Filling Approach (1992) and offers a way not only visualizing the hierachic organised tree of data but also the content of his leaf nodes. In general, you can use treemaps as a visualization method for any kind of hierarchical data. However, if you google treemap, you’ll find a lot of treemaps showing economic data like stock prices. In this field Market Watch’s Map of the Market is a great example for the two dimensions visualized by every treemap: each rectangle’s size represents market cap and each rectangle’s color represents change in market cap.

Now I asked myself wheter that visualization method could be of interest for art history: In the last years news are over and over again about record breaking prices reached for an artwork at a public auction. Such high pricing strucks not only the old masters but also works for still living artists. As you might know the prices for young artist’s paintings are often assessed by canvas size. So the question for my use-case arises: Is there also a correlation between size and hammer price of famous artworks at auctions?

I took data of the art market in 2012 based on a survey of  Statista. Because one of the TOP 20 artworks in 2012 was Jeff Koon’s Tulips (1995-2004), a sculpture, I just took the TOP 19 artworks of 2012 for my treemap use-case. I used Treemap 4.1 developed by the University of Maryland (free for non-commerical use), put in my data structured in auction house, artist’s name, size and hammer price (-> price per square inch), and got the following treemap showing the relationship between hammer price and price per square inch:

TOP19_Zuschlagpreis_PreisproQuadratzentimeter
size: hammer price (the bigger the rectangele, the higher the price); color: price per square inch (blue: low price per square inch, red: high price per square inch)

What you can read out of that visualization is:

  • Edvard Munch’s The Scream (1885) was the artwork that reached the highest hammer price,
  • Raphael’s Head of a young Apostle (1520) reached the highest price per square inch, Munch’s The Scream the second highest price per square inch, Jackson Pollock’s Number 4 (1951) the third highest price per square inch,
  • and Sotheby’s sold the most upscaled artworks including Munch’s  The Scream, Raphael’s Head of a young Apostle and Jackson Pollock’s Number 4.

Conclusion: In 2012 a lot of artworks of still living artists reached high prices at public auctions. But putting the prices in relation to their size, you can see that age dominates over size. Hence, there seems not to exist a correlation between size and hammer price but rather between hammer price and date of origin.

Veröffentlicht am Schreib einen Kommentar

What makes a helpful visualization?

Today I would like to show that visualizing results with gephi can be helpful but that not every kind of visualization algorithm implemented in gephi matters for gaining more insight. Helpfulness depends on what you would like to show! So let’s have a closer look on my use case for that!

As you may know, I write my PhD thesis about multiscreen installations. One of my favorite examples is THE HOUSE – a work of the finnish artist Eija-Liisa Ahtila that exists in two versions: a single screen version for presentation in the cinema and a multiscreen version with three projections for presentation in the gallery or museum space. Her distribution strategy opens up a lot of questions, all about what’s the difference between the two versions.

Single screen and multiscreen version have the same length (around twenty minutes). That means if you have three instead of one screen you can show three times as much. The question raises: Which of the images used in the single screen version are popping up again in the multiscreen version? And in particular: On which screen they do that?

Kanalvergleich_Übersicht_unten_zugeschnitten

Having extracted the first scene (around one minute: 1742 frames) I came to a unique result with help of Daniel Kurzawe and his application of an automatic image recognition algorithm (that is a topic for itself – paper coming soon): Most images of the single channel version reappear on the middle screen of the multiscreen version!

Using Fruchtermann-Reingold for visualizing our result I got a visualization that shows what I would like to show:

Kanalvergleich_Fruchtermann-ReingoldThe 1742 images of the single screen version (white dots) are either connected with the left, middle or right screen node (reddish dots) or aren’t connected to anything. Left, middle and right screen node grow with the number of connecting lines. The thickness of those edges again depends on how much the images resemble to each other (that’s because of our image recognition algorithm doesn’t say similiar/not similiar but gives a degree of similiartiy).

In another way also YifanHu shows the dispersion but for my flavor it is too remindful of a bacteria culture in a petri dish:
Kanalvergleich_YifanHu2

And last but not least some other visualizations I generated with gephi that are beautiful in a special sense but don’t own explanatory power because details got lost or their alignment seems too random:

 

Veröffentlicht am 2 Kommentare

Visualizing Cutting Patterns

Visualizations help people to understand complex interrelations. In two of my formerly blog posts ImagePlot: Plotting Ahtila und Plotting Ahtila (… the story continues) I showed – using ImagePlot – that the finnish artist Eija-Liisa Ahtila who produced different versions of her film THE HOUSE (2002) to explore the differences between single-screen film and multi-channel installation put most of the single-screen-material to the middle screen of the 3-channel-version. Now I would like to show another way achieving that result.

In a first step I built cutting patterns of both versions. In a second step I highlighted all shots showing Elisa, the female protagonist of THE HOUSE, and all shots which are static (that means there isn’t any activity): the first ones red, the second ones blue; white coloured shots don’t show neither Elisa nor are they static. As a result you see the following at a glance: In the 1-channel-version are a lot of shots that show Elisa and only some shots which are static, in the 3-channel-version are a lot of static shots on the left and right screen; shots that show Elisa exist mostly in the middle screen – an important outcome for understanding the steering of the beholder’s view!

1-Kanal

3-Kanal

Legende

The red and blue arrows I placed in my visualization of the 3-channel-version stand for specific interrelations between shots of the left, middle and right screen which I call Zeit-Räume and Raum-Bilder. But more about that you’ll read in my PhD thesis.

Veröffentlicht am Schreib einen Kommentar

Plotting Ahtila (the story continues)

In my last blog post I described how I explored Eija-Liisa Ahtilas 3-channel-installation THE HOUSE (2002) with ImagePlot (please see below). Ahtila is one of the artists who act not only in the art world but also in the screenland. That’s why she often produces more than one version from the same material, one multi screen version for displaying in the gallery or museum space, one single screen version for presenting in cinema. I’m sure you know I would like to get at – there’s not only a multi screen installation of THE HOUSE but also a single screen version, embedded in her portmanteau film LOVE IS A TREASURE (2002). For a better understanding of the particular potential of both presentation modes it would be interesting in which way both versions resemble each other and in which manner they differ.

Continuing my experimentation with ImagePlot, I did the same analyzes with the single screen version of THE HOUSE as I did with the single screen version before. You see the results below: The first picture shows the change of the median value (y-axis) over the film’s length (x-axis), the second shows the filled curve and the third combines both views (for this ‚combined plot‘ I manipulated the images with an image editor software).

IMAGEPLOT_INFOGRAFIK_The House_1-Kanal_klein

For comparing both versions I opposed the ‚combined plot‘ of the single screen version with the ‚combined plots‘ of the multiscreen version. In the first row of the graphic below you see the single screen version compared to the left screen of the multi screen version, in the second row you see the same compared to the middle screen and in the third compared to the right screen.

Plotting Ahtila - THE HOUSE - Comparing single and multi screen

What information can you get out of this? For me it seems that the middle screen of the multi screen version is the most similar to the single screen version. So this investigation by means of ImagePlot accounts for curator Doris Krystof’s opinion that the middle screen seems to show the main storyline:

Dabei nimmt das Bild in der Mitte insofern eine Sonderstellung ein, als man dort den Haupterzählungsstrang auszumachen meint.

– Doris Krystof (Bestandskatalog K21 Düsseldorf, Köln 2005, S. 28)