Speakers
Description
As cities around the world continue to transform through rapid urbanization, the need for public spaces that prioritize human needs and foster community engagement has become more urgent than ever. Public spaces are no longer just physical locations but are now vital to the social fabric of urban life, shaping how individuals connect, interact, and contribute to the collective urban experience. Despite this, traditional methods for assessing urban vitality remain largely confined to pedestrian counts or static measures of spatial usage, which often fail to capture the dynamic, lived experiences that define vibrant urban environments. These methods overlook the subtle, yet significant, social interactions that are at the heart of urban vitality—the moments of connection between individuals that define a space’s social energy. This research seeks to bridge that gap, offering a fresh perspective by placing pedestrian social interactions at the centre of urban vitality, offering a more nuanced and accurate measure of how public spaces truly function.
In this study, we build upon YOLOv7 to optimize and develop a social interaction recognition model specifically designed for analysing urban vitality. By integrating considerations of interaction duration and social distance, this improved model allows us to capture not only the spatial location of individuals but also their moving status（stand or move）, the number of participants, and the duration of interactions from CCTV footage. Using this detection model, we can go beyond traditional static measurements to gain a deeper understanding of the interactive dynamics of human engagement in urban environments. However, the reliance on pre-existing footage may still limit the scope of the data in some contexts. Nevertheless, the granularity and continue tracking capabilities of the model enable us to compensate for some of these limitations, offering a level of detail previously unattainable through manual observation.
To address the limitations in traditional metrics of spatial vitality, we introduce the Social Interaction Intensity Index (SIII), a novel measure that combines both the frequency and duration of interactions. This index offers a quantifiable indicator of social interaction, allowing for meaningful comparisons across different public space areas. By synthesizing these two dimensions, the SIII enables us to capture not just how often people interact, but how long those interactions last—an important aspect of urban vitality that static measures fail to account for. Through the lens of this new index, we are able to evaluate the activeness of public spaces in a more holistic way, revealing ongoing social interactions in spaces that were previously overlooked.
The study aligns with the growing emphasis on evidence-based urban planning in the era of big data. The integration of advanced computer vision technology with spatial analysis exemplifies how innovative tools can guide urban regeneration. By highlighting patterns of human interaction and their relationship with urban design, the research underscores the potential of technology-driven solutions to address contemporary challenges in urban development. Yet, like any pioneering method, this approach has its challenges. The reliance on data extracted from a specific time frame may not fully capture the broader, seasonal variations in social behaviour. To mitigate this, we have designed a robust data collection framework that includes multiple time periods across different seasons, ensuring that our findings are as comprehensive and representative as possible. In doing so, we hope to contribute to the evolving discourse on how we can design public spaces that foster connection, collaboration, and community, while addressing the evolving challenges of urban life in the 21st century and shaping the cities of the future.
| Keywords | Urban Design; Computer Vision; Social Interaction; Urban Vitality |
|---|---|
| Best Congress Paper Award | Yes |
