new icn messageflickr-free-ic3d pan white
manga.pages.1_million.Xstdev_Yentropy.jpeg_medium | by culturevis
Back to album


This visualization shows 1 million Manga pages sorted by their visual characteristics.


Software: imagej macro written by Lev Manovich


X = standard deviation

Y = entropy


this produces the following map:


horizontal dimension:

the pages on the left progressively haver fewer grey values; the pages on the right have a both black and white


vertical dimension:

pages at the bottom have only black and white

pages on the top have more grey / more detail / more realism



As we can see, the stylistic space of Manga does not have any distinct clusters. Visualization allows us to describe such a space much better than discrete linguistic categories.


The two visual features chosen for this visualization describe only some dimensions of visual style in Manga - however in terms of these dimensions, we can state this:


the concept of "style" (as a set of distinct categories used to describe a set of objects) may turn out to be meaningless


then we analyze enough objects, their variability can be better described using a continuous function


(our present research in Manga user-generated genre tags is suggesting that the same may apply for genre categories)


therefore visualization is a better language for describing cultural variability than natural languages



Note that some of the pages - such as all covers - are in color. However in order to be able to render image at this size (the original is 44,000x44,000 pixels - scaled to 10,000x10,000 for posting to Flickr), we rendered everything in grey scale.


Finally, because pages are rendered on top of each other, you don't actually see 1 million of distinct pages - rather visualization shows a distribution of all pages with typical examples appearing on the top.


2 faves
Taken on April 6, 2010