AI algorithm rates the beauty of landscapes

ga(‘send’, ‘event’, ‘Uncategorized’, ‘article’, ‘article-industry-impression’, {nonInteraction: true});

What makes a place beautiful? How does the built environment influence our lives? These are among the questions being asked by researchers at Warwick Business School’s Data Science Lab and The Alan Turing Institute. Previous studies have found that people report better health and increased happiness around what they considered to be beautiful places. At Springwise, we have already seen attempts to quantify or incorporate elements of ‘scenicness.’ These have included a mapping tool that takes users to the most beautiful locations, and proposals to move traffic underground to make cities more scenic.

In an attempt to quantify what might be considered scenic, the researchers used more than 200,000 images of places that had been rated as highly scenic by users of the lab’s Scenic-or-Not website. The researchers analysed the images using the MIT Places Convolutional Neural Network to find what attributes, such as ‘trees’, ‘mountain’, and ‘highway’, corresponded to high and low scenic ratings. The researchers then adapted the algorithm to rate the beauty of new locations. They found that participants associated features such as ‘valley’, ‘coast’, ‘mountain’ and ‘trees’ with higher ‘scenicness’. On the other hand, some constructed elements also tended to improve scores. These included not only historical architecture such as ‘church’ and ‘castle’, but structures such as ‘viaduct’ and ‘aqueduct’. Interestingly, large areas of green space such as ‘grass’ and ‘athletic field’ led to lower ratings of ‘scenicness’.

According to lead researcher Chanuki Seresinhe, “It appears that the old adage ‘natural is beautiful’ seems to be incomplete: flat and uninteresting green spaces are not necessarily beautiful, while characterful buildings and stunning architectural features can improve the beauty of a scene.”

Takeaway: Researchers hope that the algorithm could help inform future planning decisions. Such developments aim to improve the sustainability of new buildings and the wellbeing of residents. How else could deep learning inform other types of planning decision one day?

Website: www.datasciencelab.co.uk
Email: [email protected]

Source: New feed 1