Saturday, July 2, 2022
HomeMachine learning & AISocial media and AI can measure the aesthetic quality of landscapes

Social media and AI can measure the aesthetic quality of landscapes

Social media and AI can measure the aesthetic quality of landscapes

The model with Flickr and AI generates an overall measure of landscape aesthetic quality for each area. Courtesy: Ilan Hanga (Photo by Sergio and Graeme Charchard (CC-2.0)

Helps to inform public environmental policy to measure the beauty and well-being of an ecosystem. Scientists at EPFL and the University of Wageningen in the Netherlands have developed a new formatting approach to ecosystem assessment based on in-depth learning and millions of Flickr photos.

The extent to which we enjoy outdoor activities such as mountaineering, running or boating in the mountains largely depends on the beauty of the surrounding ecosystem. For example, landscapes with blue-blue seas, yellow and lavender-covered mountains or rocky outcrops can be good for our physical and mental health. This sense of well-being is one of the factors examined in Ecosystem Services (ES) assessments, which measure the contribution of landscapes to the well-being of people for informing environmental policies. To support these assessments, a team of scientists from EPFL and the University of Wageningen have developed a new model that incorporates artificial intelligence into human aesthetic enjoyment. Their model, which draws on over nine million pictures of British landscapes posted on Flickr, can be easily reproduced on a large scale as it uses data on social media. It is this model that first includes an understanding of how individuals value landscapes on a large scale while remaining as accurate as they are today. Found in their research Scientific reports, A natural expression.

To improve their model, scientists have developed an in-depth learning algorithm for over 200,000 photographs of landscapes in Great Britain using a landscape or non-landscape database. These photographs, consisting of a database of geographical representations of Great Britain, were rated by a population survey in terms of their aesthetic quality or “scenery.” This approach enabled scientists to incorporate evidence of missing elements from traditional large-scale ES assessments — how to enjoy landscapes privately. The research team then implemented the neural network-based in-depth study model on more than nine million Flickr images, and linked other AI-based models to predictions about their beauty. Finally, they compared the results of their model with those of a more traditional environmental indicator-based model.

A more accurate model

Scientists have compiled their findings using frequency symbols that show the color of the scene on maps of Great Britain. Both models found that the Snowdonia National Park in Wales, the Lake District of England and the Highlands of Scotland were particularly high aesthetic value and well-being areas. “The results of the two models are roughly identical at a resolution of 5 km,” said Davis Tuya, an assistant professor at the EPFL’s Environmental Computer Science and Earth Observation Laboratory. They both clearly recognized that urban areas like London and Glasgow were not attractive. But at a resolution of 500 m² inconsistencies arise and the Flickr mode is more accurate. For example, Greater London, Richmond Park and Heathrow Airport are predicted to be the most scenic areas in the traditional format, while Flickr mode categorizes them as the most inaccessible.

An entirely new way of assessing the environment and how people interact with it

Thanks to a combination of social media and in-depth learning, it is also possible to estimate how people’s appreciation of the aesthetic quality of a particular landscape changes over time to the model of scientists. In an additional experiment, the research team looked at the UK’s natural park areas, Lake District, Pembrokeshire Beach in Wales and Keongorms in Scotland for their unique beauty. This experiment enabled them to study how aesthetic factors relate to the seasons. The “snow” quality, for example, coincides with the weather forecast for that period – the new model accurately showed that the winter of 2009-10 was particularly snowy. Scientists have even found that the prevalence of snow increases over the weekend, with people more likely to view snowy landscapes, while the prevalence of “asphalt” remained relatively low throughout the week. “This implies that social media-based data use provides a collection of information about the state of the environment and how people interact with it,” Tuya said. “Information like this has never been obtained with such high accuracy before.”

Ilan Hinga, Ph.D. A student at the University of Wageningen says: “It is not easy to measure on a large scale how the aesthetic quality of the landscape contributes to the well-being of our people. It is to explore whether this model can be applied to other countries, considering how different the landscapes and cultures may be. Scientists will need to find a way to train AI algorithms using locally relevant criteria. Projects are already underway in the Netherlands, Spa Spain and other European countries to support environmental protection policies across Europe.

The cost of wind farm development from the landmark study is shown in the picture

More info:
Ilan Hinga & Al, captures the aesthetic qualities of landscape through social media and deep learning, Scientific reports (2021). DOI: 10.1038 / s41598-021-99282-0

Provided by Ecole Polytechnique Federale de Lausanne

Excerpt: Measure the aesthetic quality of landscapes on social media and AI (2021, October 15) October 17, 2021 at Retrieved.

This document is subject to copyright. No part of it may be reproduced without written permission, except in fairness for personal study or research. Content is provided for informational purposes only.




Please enter your comment!
Please enter your name here

Most Popular

Recent Comments