Although beauty is often understood to be a subjective concept, harnessing the insights from tens of thousands of people through machine learning allows us to distil a shared, collective understanding of beauty.

Back in 2017, we developed a methodology that was applied to England and published in the peer-reviewed open-access scientific journal, Royal Society Open Science. 

Recognizing the potential of Convolutional Neural Networks (CNNs) in image tasks, we employed them to evaluate the beauty of various locations.  Our goal was to have the CNN evaluate the aesthetic appeal of the environment captured in the photos, not just the photos themselves.

To enhance our model's performance, we utilised a method known as transfer learning to modify an established model, the MIT Places365 CNN, to better understand and rate scenic beauty.

In practical tests on London imagery, our model astutely identified the beauty in natural settings like parks and was also adept at recognizing the splendor of architectural landmarks.

In summary, our 2017 research demonstrated that machines can indeed play a role in quantifying the beauty of our surroundings. For more information on this methodology, refer to our study: Using Deep Learning to Quantify the Beauty of Outdoor Places.