What is it about?
This research uses vision language models (CLIP) to develop a new indicator of perceived walkability that takes into account physical features (greenery, visible sky, buildings etc.) alongside subjective factors (safety, attractiveness). The subjective factors are assessed using CLIP, mimicing a human's perception of the urban environment. The results provide more nuanced assessments than scores based on phyiscal features alone.
Featured Image
Photo by Jace & Afsoon on Unsplash
Why is it important?
The research shows how visual language models can be used to percieve images in a more human manner, approximating a human's subjective judgement of the urban environment.
Perspectives
This research is a first step in understanding how large language models, including vision language models, may improve on traditional deep learning architectures in urban analytics applications.
James Haworth
University College London
Read the Original
This page is a summary of: A New Approach to Assessing Perceived Walkability, November 2023, ACM (Association for Computing Machinery),
DOI: 10.1145/3615888.3627811.
You can read the full text:
Contributors
The following have contributed to this page
