Using street view images and visual LLMs to predict heritage values for governance support: Risks, ethics, and policy implications
Researchers have developed a method using visual large language models (LLMs) to identify buildings with heritage value in Sweden, analyzing over 154,000 street view images. This approach aims to support Swedish authorities in creating National Building Renovation Plans, as required by the EU's Energy Performance of Buildings Directive. The study highlights potential risks associated with using LLM-generated data for governance, including issues of transparency and error detection. AI
IMPACT Demonstrates a novel application of visual LLMs for identifying heritage buildings, potentially aiding policy implementation and urban planning.