urban design metrics

[tobiaskauer]

The paper draft suggests five urban design metrics:

• Walkability
• Visual Complexity
• Greenspace
• Landmarks
• Openness

Questions:

• What are heuristics to calculate each dimension?
• How are the LAWN-categories used to calculate metrics?
  • Landmark (translates to landmark?)
  • Architecture
  • Walkablility (translates to walkability?)
  • Natural
• Why and how do these metrics deviate from the initial suggestion in the urban design metrics book?
  • imageability
  • enclosure
  • human scale
  • transparency
  • complexity
  • (coherence)
  • (legibility)
  • (linkage)

[tobiaskauer]

(copy paste from slack)

So there are seperate ways to generate the scores: All are approximate offcourse becasue our model does not spit out any particular score, The process for each metric i think would be the following:

openness and green spaces

• I have a trend graph that shows abinned plot of how green cover and openness (area of sky area) is linked with beauty.
• i will use these ranges as scores and if an image falls in that range i will give it that score
• so for openness, if image has more than 40% sky pixels it has a score of 5, 0 - 0.1 score of 1 and so on
• the findings suggest less open => more beauty
• similarly for tree cover

Landmarks & Walkability

• for Landmarks , i am thinking of looking at the top 5 scene types given to an image, and how many of them fall in the category of Landmark would be the score
• same for Walkablility

Visual Complexity

• visual complexity is a bit convoluted - lets drop that for now from your visualization
• the way i am calculating it is through the distribution of each segment across image:
• so more close to uniform the distribution, the more the visual complexity. so if there are equal number of cars, pedestrians, buildings, trees then that is more complex. i am still not sure about this, this was miriams idea. you will always have peaks, e.g. sky pixels will pretty much always over power Fence pixels. so you will never get uniform distribution of pixels