The German artists Bernd and Hilla Becher, who began working together in 1959 and married in 1961, are best known for their “typologies”—grids of black-and-white photographs of variant examples of a single type of industrial structure. 

Here, machine learning is used to visualize the typologies documented by the Bechers.

The rigorous frontality of the individual images gives them the simplicity of diagrams, while their density of detail offers encyclopedic richness. At each site the Bechers also created overall landscape views of the entire plant, which set the structures in their context and show how they relate to each other. 

The analysis of the Bechers’ images attempt to quantify the quality of frontality, simplicity, and density.
Frontality through the total area of transparent elements; Simplicity interpreted as the number of asymmetrical lines; and density by finding the total area of distinct regions found.

To create these works, the artists traveled to large mines and steel mills, and systematically photographed the major structures, such as the winding towers that haul coal and iron ore to the surface and the blast furnaces that transform the ore into metal. 

To account for variations between photographs and typologies, an evolutionary solver is run to find the optimal setting for each photograph.

The typologies emulate the clarity of an engineer’s drawing, while the landscapes evoke the experience of a particular place. The exhibition presents these two formats together; because they lie at the polar extremes of photographic description, each underscores the creative potential of the other.

The geometric representation of the typologies defined by the Bechers hint at qualitative descriptors. Proximity suggests similarity. An isolated typology suggest highly distinguishable types whereas intersection hints at marked similarities. Likewise, a large typology suggests variation and a smaller geometry hints at a well defined typology.