{"id":30866,"date":"2025-06-11T18:11:21","date_gmt":"2025-06-12T01:11:21","guid":{"rendered":"https:\/\/essential.construction\/news\/how-machine-learning-can-help-with-urban-development\/"},"modified":"2025-06-11T18:11:21","modified_gmt":"2025-06-12T01:11:21","slug":"how-machine-learning-can-help-with-urban-development","status":"publish","type":"post","link":"https:\/\/essential.construction\/news\/how-machine-learning-can-help-with-urban-development\/","title":{"rendered":"How Machine Learning Can Help with Urban Development"},"content":{"rendered":"

[ad_1]
\n<\/p>\n

\n

An
\nexperimentation project has demonstrated the capabilities of machine learning
\nin urban development. It used images as a starting point and came up with
\ninteresting and useful applications.<\/p>\n

<\/p>\n

\n
\"-\"\"-\"<\/figure>\n<\/div>\n

\u201cI read data
\nscience papers on how machine vision algorithms can be used with satellite
\nimagery. I immediately saw a connection to what we had been doing,\u201d Antti
\nKauppi<\/strong>, architect at Arkkitehdit Sankari, explains. \u201cMost people associate
\nimage recognition with Google\u2019s visual searches. Google can distinguish whether
\na photo shows a cat or another animal, for example. We went a step further.\u201d<\/p>\n

An Experiment with Open Urban Imagery<\/h2>\n

Arkkitehdit
\nSankari Oy, a Finnish architectural design firm began the experimentation
\nproject CityCNN<\/a> in May 2018. It received funding from KIRA-digi, the
\nFinnish government\u2019s digitalization program for the built environment. CityCNN
\nexplored the possibilities of using machine learning and open data for urban
\ndevelopment.<\/p>\n

Kauppi
\ncollected data from Espoo, Finland\u2019s second-largest city on the outskirts of
\nHelsinki. He created a piece of Python software to retrieve data from the
\ncity\u2019s server.<\/p>\n

Kauppi has
\ndone programming for many years but does not consider himself an AI expert: \u201cMy
\nspecial skill is to apply the technology to our business. It\u2019s much like using
\nPhotoshop. You can use it successfully even if you don\u2019t master its inner
\nworkings.\u201d<\/p>\n

Using Competing Neural Networks<\/h2>\n

The
\nexperiment used so-called generative adversarial networks (GANs). It is a
\nmachine learning technique in which you match two neural networks against each
\nother.<\/p>\n

The first
\nnetwork, the generator, creates new images. The second network, the
\ndiscriminator, uses real images and takes in the newly generated images. It
\nevaluates whether an image is real or generated. In repeating this process over
\nand over again, the generator and the discriminator become more accurate and,
\nas a result, the generated images improve.<\/p>\n

Conditional adversarial
\nnetworks used in CityCNN are an extension of the basic technique. They are
\ntrained with image pairs. They can create photographic images from line
\ndrawings or convert an impressionist painting into a photograph, for example.<\/p>\n

\"-\"\"-\"
The conditional adversarial network as used in CityCNN<\/figcaption><\/figure>\n

The Experiments<\/h2>\n

\u201cWe use
\nAmazon\u2019s cloud for computational power. I\u2019ve trained the network with hundreds
\nof image pairs,\u201d Kauppi explains. He shows examples of aerial photos that the
\nnetwork has converted into city plans. By reversing the neural networks, the
\nsystem created aerial illustrations using city plans.<\/p>\n

Another
\nCityCNN application marked buildings in a satellite image with a color. Kauppi
\nlikens it to a five-year-old who\u2019s given crayons and told to color all the
\nbuildings. Reversing the action, the neural network can create satellite images
\nfrom building masses, automaticallly adding roads and streets and even parking
\nlots for larger buildings. It has learned what the landscape in Espoo looks
\nlike and mimics it fairly accurately.<\/p>\n

\"-\"\"-\"
From density map to satellite image<\/figcaption><\/figure>\n

Kauppi gives
\na live demonstration of the network\u2019s capability. He draws rectangles in an
\nempty window and the machine creates a counterpart in another. When he draws a
\nsmall rectangle in the middle, the machine treats it as a house in the middle
\nof a tree-covered area. A long narrow line ends up being a street with houses
\non both sides. When Kauppi draws large building masses, the network sees them
\nas an industrial complex or warehouse and creates adjoining large parking areas
\nautomatically.<\/p>\n

All the
\nvisualizations are algorithmic and based on mathematics. The machine does not
\n\u201cunderstand\u201d the context. However, its behavior looks disturbingly human.<\/p>\n

Practical Applications<\/h2>\n

An
\napplication that may have a practical use right away is a tool that identifies
\nareas for potential supplementary development. Kauppi had taken aerial photos
\nof Espoo\u2019s residential areas and used a paint program to mark spaces he deemed
\nsuitable for infill. Using 500 image pairs as training material, he taught the
\nnetwork to do the same to any satellite image. This way, the machine could
\nquickly spot all the potential areas for supplementary development.
\nIntelligently, it did not flag parks or woods.<\/p>\n

\u201cIf we gave
\n5,000 image pairs to experts and had them mark meaningful things on the images,
\nthe network would learn how to do the same on a national level,\u201d Kauppi
\nenvisions. \u201cThat sounds like a lot of work, but it would take about 10 days and
\na few thousand euros for the computing, which is reasonable. After the initial
\ntraining, the network can generate new images in milliseconds.\u201d<\/p>\n

\"-\"\"-\"
Generated bitmap compared to a real one<\/figcaption><\/figure>\n

If you
\ncombine image data with other types of urban data, new applications emerge.
\nKauppi mentions Mapita\u2019s Maptionnaire app, which allows citizens to give
\nlocational feedback on their urban experiences. If people tag certain areas as
\nbeing unsafe or pleasant, a machine learning algorithm could automatically
\nlocate other similar places to help with city plans of the future.<\/p>\n

\u201cNow that
\nwe\u2019ve completed this experiment satisfactorily, we\u2019ll report the results and
\nshare our experiences openly,\u201d says Kauppi. \u201cWe\u2019re happy to discuss how to
\ndevelop these ideas further.\u201d<\/p>\n

You can email Antti Kauppi at

[ad_2]
\n
This article was originally posted at Source link <\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

[ad_1] An experimentation project has demonstrated the capabilities of machine learning in urban development. It used images as a starting … Read more<\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1062,1066],"tags":[1168],"class_list":["post-30866","post","type-post","status-publish","format-standard","hentry","category-aec-business","category-all-posts","tag-urban-development","generate-columns","tablet-grid-50","mobile-grid-100","grid-parent","grid-33"],"_links":{"self":[{"href":"https:\/\/essential.construction\/news\/wp-json\/wp\/v2\/posts\/30866","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/essential.construction\/news\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/essential.construction\/news\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/essential.construction\/news\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/essential.construction\/news\/wp-json\/wp\/v2\/comments?post=30866"}],"version-history":[{"count":0,"href":"https:\/\/essential.construction\/news\/wp-json\/wp\/v2\/posts\/30866\/revisions"}],"wp:attachment":[{"href":"https:\/\/essential.construction\/news\/wp-json\/wp\/v2\/media?parent=30866"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/essential.construction\/news\/wp-json\/wp\/v2\/categories?post=30866"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/essential.construction\/news\/wp-json\/wp\/v2\/tags?post=30866"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}