Science & technology | Artificial intelligence

Computer analysis of what is scenic may help town planners

Admiring the scenery

Can you tell a green field from a cold steel rail?

BEAUTY, proverbially, is in the eye of the beholder. But surroundings matter. A paper published two years ago in Nature found a correlation between people’s sense of well-being and the “scenicness” of where they lived. The paper’s authors measured scenicness by asking volunteers to play an online game called Scenic-or-Not, which invites participants to look at photographs of neighbourhoods and rate their scenic value on a scale of one to ten.

The correlation, the paper’s authors found, held true whether a neighbourhood was urban, suburban or rural. It bore no relation to respondents’ social and economic status. Nor did levels of air pollution have any influence on it. The authors also discovered that differences in respondents’ self-reported health were better explained by the scenicness of where those respondents lived than by the amount of green space around them.

Pinning down what scenicness actually is, though, has always been a frustrating exercise for scientific types. The team behind that Nature paper, Chanuki Seresinhe and her colleagues at Warwick Business School, have nevertheless decided to have a go. And they think they have succeeded. As they report in Royal Society Open Science, they have adapted a computer program called Places to recognise beautiful landscapes, whether natural or artificial, using the criteria that a human beholder would employ.

Places is a convolutional neural network (CNN), a type of program that can learn to recognise features in sets of data, such as images, presented to it. CNNs often form the basis of face-recognition software. Places, though, as its name suggests, is optimised to recognise geographical features. Ms Seresinhe and her team taught the program to identify such things as mountains, beaches and fields, and various sorts of buildings, in pictures presented to it. Having done so, they then fed it with 200,000 photos that had been assessed by players of Scenic-or-Not. The program’s task was to work out, by analysing each photograph’s features in the context of its Scenic-or-Not ratings, what it is that makes a landscape scenic.

Most of the results are not surprising. Lakes and horizons scored well. So did valleys and snowy mountains. In artificial landscapes castles, churches and cottages were seen as scenic. Hospitals, garages and motels not so much. Ms Seresinhe’s analysis did, however, confirm one important but non-obvious finding from her previous study. Green spaces are not, in and of themselves, scenic. To be so they need to involve contours and trees.

This observation plays into an idea promulgated 30 years ago by Edward Wilson, an evolutionary biologist at Harvard University. He suggested that the sorts of landscapes people prefer—and which they sculpt their parks and gardens to resemble—are those that echo the African savannahs in which Homo sapiens evolved. Gently undulating ground with a mixture of trees, shrubs and open spaces, in other words (though, ideally, without the accompanying dangerous wild animals).

In particular, the parks laid out by 18th-century European magnates often fit these criteria. And those parks are also replete with follies—small buildings or imitation ruins of the sort Ms Seresinhe’s work suggests people generally find scenic, too. There is a message here for town planners. Less grass and more trees and bushes would be welcome. And perhaps, also, the odd deliberate folly dotted around, as opposed to the accidental follies that make up so much of modern architecture.

This article appeared in the Science & technology section of the print edition under the headline "Admiring the scenery"

Together, technology and teachers can revamp schools

From the July 22nd 2017 edition

Discover stories from this section and more in the list of contents

Explore the edition

Discover more

Antarctica, Earth’s largest refrigerator, is defrosting

The world must pay more attention to its southern pole

Killer whales deploy brutal, co-ordinated attacks when hunting

Their techniques are passed down through the generations


A new generation of music-making algorithms is here

Their most useful application may lie in helping human composers