Identify Pedestrian Social Interactions: A Computer Vision Approach to Social Interactions and Urban Vitality (CUPUM 2025)

As cities urbanize rapidly, the need for public spaces that foster connection and community is urgent. Traditional methods of assessing urban vitality, like pedestrian counts, overlook dynamic social interactions critical to vibrant environments. This research addresses that gap by developing a social interaction recognition model based on YOLOv7, capturing spatial locations, interaction duration, social distance, and movement from CCTV footage. To evaluate urban vitality, we introduce the Social Interaction Intensity Index (SIII), combining interaction frequency and duration for a comprehensive measure of public space activeness. By integrating computer vision with spatial analysis, this study reveals how social interactions reflect urban vitality, advancing evidence-based planning and contributing to designing more connected, vibrant cities.