Coming to terms with ambiguous asset dimensions: quality, beauty and uniqueness in the built environment
Coming to terms with ambiguous asset dimensions: quality, beauty and uniqueness in the built environment
In this project we will first train a computer vision model from ground up. The resulting machine learning model will be the first to be fully optimised for the built environment, offering an alternative to existing transfer-learning approaches and general purpose computer vision models. Second, we will quantify and classify hard to measure or ambiguous visual building attributes such as asset uniqueness, perceived quality of materials and designs, or architectural beauty and character of buildings. Many of these perceptions are inherently idiosyncratic, which is why we develop a personalised recommendation engine for residential real estate: “Based on earlier choices, we believe you might find this building appealing...”