How we used Google Image Search data to observe Lombardia’s landscape
The painting Landscape with Footbridge by Albrecht Altdorfer is one of the first example of pure landscape art (representing nature without humans). Since then landscapes have always been mainly background material for religious themed paintings. After this early attempt of representing the natural environment with humans out of the picture, man-made reproductions of the natural landscape have come a long way. Together with that, the actual impact of man over its environment has changed along. We now live in the world of indoor ski slopes in the desert, cellphones towers disguised as palmtrees and german tropical islands. The natural environment has been so variously changed, represented, copied, referenced, camouflaged and manufactured that the boundary between natural and artificial is now more than ever blurred and unstable.
This experiment is an attempt at using found digital material (i.e. images from Google Image Search) to map and visually compare how different countries perceive, represent and imagine the landscape of Lombardia, a region in Northern Italy. Which colors, shapes, textures and materials resonate the most outside Italy? Is there a shared stereotyped landscape? Or are there multiple changing Lombardies to be discovered, as as long as the point of view changes?
Web scraping. The local domain of Google Image Search of 8 countries (United States, India, Brazil, Indonesia, Germany, Netherlands, France, Spain) has been queried with the term [Lombardia], translated in the local language (language versions of the Wikipedia article Lombardia have been used as a reference). The search has been done with the photo type filter active, in order to exclude maps and graphics. Images returned from each search have been then saved locally using the Mozilla Firefox add-on DownThemAll.
Filtering. From each national collection, images have been manually tagged in order to retain only those ones depicting a landscape. This process has been particularly critical as the definition of what constitutes a landscape could heavily affect the selection of images used and therefore resulting in very different outputs. As the focus of the experiment was on comparing landscapes’ visual features (colors, textures, lines and materials), only images containing an horizon line were retained. Close-ups of objects, mugshot style portraits, bird-eye views and cartographic maps have been discarded.
Remix as knowledge making strategy. Eli Horvat defines remix as the practice of appropriating pre-existing film footage in order to denature, detourn, or re-contextualise images by inscribing new meanings onto materials through creative montage. The method used for this experiment draws upon the idea that remixing existing (visual) material can actually serve as a mean to acquire new knowledge from and about the material itself. Remix is employed here as a knowledge-making activity as well as a tool to experience information simultaneously.
Sampling. Some authors already mention the step of sampling in the definition of remix itself. Eduardo Navas describes remix as the activity of taking samples from pre-existing materials to combine them into new forms. It goes without saying that the sampling process was a critical step of the experiment. Lev Manovich clearly states that sampling cannot be treated as a mechanical process but represents instead a creative strategy that can be applied to any dimension of the media data. The choice of what and how to sample can produce very different outputs and therefore enable diverse observations. The sampling method chosen here aims at extracting slices big enough to still show the textures and the materials depicted in the images (and not just the colors) but small enough to be easily re-organized later in a larger composition. After resizing all images to the same height, we extract two slices of the same height from each image: one from the bottom part and one from the upper one.
Measuring. Measuring dominant colors for each image served the purpose of adding an additional data point for sorting each collection of images. The Color Thief library has been used to extract dominant Hue, Saturation and Lightness (HSL) values from each slice.
Layout. The main goal of the experiment was to visually compare textures and colors of Lombardia as they are perceived by different countries. We designed the final output trying to enable macro comparisons between countries, pattern detection within a specific country and detailed observations of each slice. We created an image-montage composition for each country, sorting slices from the top to the bottom of the page, according to their dominant hue. The eight produced collages have been juxtaposed side by side.
Looking at the final output, there is a difference in the amount of landscape images in each collection. The query in Spanish [Lombardía], for example, returned roughly half of the images obtained with the query in German [Lombardei]. Looking at the overall color composition, no particular difference between countries can be found. There is a clear predominance of blue shades across all countries and an even clearer lack of red shades. All countries composites can easily be divided into three big sectors.
A grey and brown group of images at the bottom, constituted by infrastructures and architectural elements (with the Duomo of Milan peculiar texture being the more prominent one). A middle stripe where brown shades slowly turn into the greens of landfills and mountains, and a light blue one containing skies and snowed mountain tops.
To a closer observation, some country-specific behaviors emerge: India and Spain present a slightly smaller amount of green shades, while Netherlands and Germany seem to contain more blue shades with respect to other countries. These country-specific features suggest to take a closer look to particular sections of the composites and reveal unexpected results. The prominence of blue in the Dutch and German landscape is determined by the presence of a good number of slices depicting water from swimming pools: reconciling those slices with their original images and using Google Image Search to locate them online, led us to a panorama of private luxurious villas for sale and rent, all of them in the area of Como lake. For Dutch and German users, Lombardia is (also) a luxury retreat spot and a real estate investment opportunity.
The final output, showcased inside the Milano Expo 2015 exhibition Lombardies, offers a first entry point for the description of Lombardia’s landscape as mediated by a country-specific digital device (Google local domains) and suggests possible directions for in-depth analyses taking into account image contents.
Describing the cognitive relationship between men and his environment, XIX century geographer Alexander von Humboldt, defines three different separate stages: at first, an irrational sensation of awe in which the greatness and beauty of natural elements are experienced as a whole totality (Eindruck). After a first initial emotional drive, nature is dissected and translated into a collections of elements, scientifically and rationally described (Einsicht). In the last phase of the process (Zusammenhang), the totality of the environment is restored, this time at a scientific level, and the elements previously analysed as single units are put back together thanks to the study of their relationships. Humboldt clearly states that any kind of scientific knowledge about the environment cannot be produced without that first irrational impression of totality.
The method and output described here can serve a wide array of purposes and enable very different level of interpretations. But first of all the variety of those interpretations can provide that very first experience of totality which can be a valid entry point into the scientific study of a geographic area mediated by digital technology.
Farinelli, Franco. “The loop of the landscape”. (2012). Retrieved November 29, 2015, from http://www.comune.modena.it/galleria/press-room/press-archive/2012/vittorio-corsini.-amongst-voices-papers-brambles-and-nocturnes/the-loop-of-the-landscape-by-franco-farinelli-1/view?set_language=en
Horwatt, Eli. “A taxonomy of digital video remixing: Contemporary found footage practice on the Internet.” Cultural borrowings: Appropriation, reworking, transformation (2009): 76–91.
Manovich, Lev. “Museum Without Walls, Art History Without Names: Visualization Methods for Humanities and Media Studies.” (2013): 253–278.
Navas, Eduardo. “ Remix Theory — Remix Defined.” Retrieved November 28, 2015, from http://remixtheory.net/?page_id=3
O’Neil, Jamie. “Mix/Remix as Epistemology: The Implications of the Metamedium, Digital Media.” McLuhan Remix. np 16 (2006).
Wolff, Rachel. “Cut-and-paste culture: the new collage.” Retrieved November 30, 2015, from http://www.artnews.com/2013/12/12/the-new-collage/