CommuniWave Model Quantifies Informal Urban Behaviors from Video

Urban planning often follows a top-down approach, frequently clashing with the actual, often informal, behaviors of residents. A significant challenge for urban managers and designers is the lack of effective metrics to quantify these temporary informal behaviors, which are crucial for enhancing community resilience and functional attributes. To address this, a new machine learning model named CommuniWave has been introduced. CommuniWave is designed for efficient detection and quantification of the "Degree of Informal Behavior" (DIB) within urban communities, utilizing street video footage. The model integrates three core components: a Behavior Capture Net (BCN) built on mmaction2 for action recognition, a custom-developed YOLOv10 model (YLX) for object detection, and a Random Forest-based Behavior Eval Model (BEM) to synthesize the data. By generating DIB fluctuation charts from continuous video streams, CommuniWave enables dynamic monitoring of urban spaces. This capability provides urban managers with refined data-driven insights, supporting more adaptive decision-making to improve the overall resilience and functionality of communities, bridging the gap between formal planning and actual usage patterns.