Category: Projects

Looking Down (live link) catalogs aerial images taken at a set of points in Prospect Park. The images are positioned relative to these points. On hover, an aerial image flips to an image of the sky taken at the same point.

“Looking Down (015)” is coupled with “Looking Up (012)”. In juxtaposing photographs of the sky with GPS accuracy readings, “Looking Up” described the relationship between tree canopy, open space, and accuracy. By exchanging one half of the pair, accuracy readings, for aerial images of the same location, “Looking Down” asks: does the sky match the ground? Does the ground provide context for the sky or vice versa? Does the human-scale of the sky-facing image alienate the otherwise familiar, but detached aerial perspective? Does the sky or ground seem more continuous than the other?

Field documentation methods are detailed in the post, “Field Data Collection (001): Prospect Park Walk“.

Next Steps:
– Pixelate images according to the accuracy reading on hover.
– Layout images according to recorded GPS coordinates at each point rather than the idealized grid. Some images will overlap and some will misalign.
– On hover, randomly flip half of all tiles, so that you never see all of the ground or the sky.
– On refresh, start with the aerial; refresh again, start with the sky.

Compare to:
– Looking Up (012)

Subway Stream (live link) presents the advertisements of two subway stations as images whizzing by like passing trains. As more stations are added: How is a particular subway station identifiable from its set of ads? What ads repeat across stations and within a station? Where is there vacancy?

Photographs were taken the morning of Tuesday, January 23rd at the Seventh Avenue B/Q stop in Brooklyn and the Prince Street 6 stop in Manhattan. The automatic movement of the images is achieved through an interval function calling scrollLeft() every 50ms and sliding the images over 200px.

Next steps:

• Add additional subway stations in different parts of the city.
• Clean up code: make a class from which each station is instantiated, include direction, speed and extents as properties.

A Walk Three Ways (013) is a triptych of a single walk through Prospect Park. Users can scrub through an altitude graph, aerial image, and plan, all of which describe the same route.

The linearity of the altitude graph conceals the circuitousness of the path that it describes; whereas, in plan, the tangled path obscures any notion of topographic variation. The aerial image, while still planimetric, begins to illustrate the conditions that lead to the variation in both dimensions. The interaction links these differences: hovering over one aligns the others to the same position.

The walk followed grid-points spaced at 200m. The data collection process is detailed in the post “Field Data Collection (001): Prospect Park Walk“. The path was recorded using a Bad Elf GPS data logger and processed in QGIS, from which it was exported to GEOJSON. The altitude graph was built using d3.js, while the map-based representations used Mapbox GL.

Next Steps:

• Adjust zoom levels such that each visible extent is the same, and try another in which they are distinctly different.
• Smooth out the path by culling points recorded when we were stationary.
• Remove panning from aerial representation, or add interaction which snaps to the closest point on the path.
• Adjust x-axis of altitude graph to reflect time more accurately.
• Add hover interaction to plan representation.

 

Looking Up (live link) is a collection of photos of the sky positioned relative to where they were taken. On hover, the GPS accuracy at each point is displayed. Both the accuracy readings and photos offer an oblique description of the tree canopy and open space in Prospect Park.

The photos were collected as part of a larger adventure through Prospect Park, detailed in the post “Field Data Collection (001): Prospect Park Walk“. Each photos was taken while facing north, and taken as a whole, the set describes the duration of the data collection.

Next Steps:

• Show corresponding aerial image on hover.
• Pixelate images according to the accuracy reading on hover.
• Experiment with different sampling densities.
• Layout images according to recorded GPS coordinates at each point rather than the idealized grid. Some images will overlap and some will misalign.
• Re-export images to match web standards and reduce loading time.

 

“Dominant Forms: Areas” is a figure-ground representation of all the buildings in New York City. On hover, the figure-ground is filtered to show only buildings that have a similar footprint area. This categorization asks: which neighborhoods have same-sized buildings and which have a variety of footprints? Where are the largest footprints and where are the smallest? How are these distributions different from one borough to another?

These patterns are typically obscured by the visual density of aerial photography or multi-layered drawings. By isolating a category, such as footprints, differences within the category can be found. Compare to “Points & Polygons (004)“, where difference is identified between categories.

The map purposely restricts zooming to prevent an all-encompassing view of the city and to individuate building footprints. Thus, when the buildings are isolated by size, the particular shapes are more apparent than if zoomed out, and adjacencies are more apparent than if zoomed in.

Building areas were separated into eight buckets:

• under 500 sq ft
• between 500 and 2,000 sq ft
• between 2,000 and 5,000 sq ft
• between 5,000 and 10,000 sq ft
• between 10,000 and 25,000 sq ft
• between 25,000 and 50,000 sq ft
• between 50,000 and 100,000 sq ft
• over 100,000 sq ft

These buckets were determined by trial-and-error, however, cluster analysis will be used to develop more nuanced categorizations.

Next Steps:

• Use cluster analysis, such as k-means, to better identify building size groupings
• Upload all footprints to Mapbox studio to improve performance
• Add debounce feature on hover

Compare to:

• Points & Polygons (004)