Evaluate Spatial Equity in New York City’s Vibrant Street Programs with Local Lightweight Vision LLM (Working Paper)

Apr 1, 2025·
Yang Yang
· 3 min read
Abstract
This study evaluates the spatial equity of New York City’s vibrant street programs—such as Open Streets and Open Restaurants—by examining whether planning and implementation have prioritized historically underserved neighborhoods. We introduce a scalable, privacy-preserving methodology using a local lightweight vision large language model (vLLM), Gemma 3, to detect program implementation by comparing pre- and post-pandemic Google Street View images across thousands of locations. Findings reveal major implementation gaps, especially in the outer boroughs, despite equitable intentions in planning. This study underscores the need for better resource allocation and community support to ensure fair and effective delivery of urban public space programs.
Type

The COVID-19 pandemic dramatically reshaped urban life, particularly in densely populated cities like New York City (NYC) (Bereitschaft & Scheller, 2020). Social distancing measures, lockdowns, and the shift to remote work have greatly reduced face-to-face interactions, altering the social fabric and vitality of communities (Bojczuk et al., 2024). As cities recover from the pandemic, there is an urgent need to reimagine and revitalize street spaces to foster renewed social interaction and rebuild community connections for the people living there (Thorpe, 2021). In order to “reclaim the streets of New York City for its people,” the city government has launched various vibrant street programs such as Open Restaurants and Open Streets, following the complete street concepts in its latest Street Plan (NYCDOT, 2021).

These programs are designed to create safe, accessible, and equitable street space for all users, regardless of age, ability, or mode of transportation. However, concerns about spatial equity persist, with news reports suggesting significant misalignment between planned and implemented programs in disadvantaged neighborhoods (Hirsch, 2021). It was estimated that over 84% of the planned open street programs in the Bronx were not implemented due to low participation and suboptimal street infrastructure. The phenomenon reflects that an equitable plan may not necessarily lead to an equitable city, thereby affecting the long-term value of these programs.

This study evaluates the spatial equity of NYC’s vibrant street programs in both planning and implementation, addressing two questions:

(1) Were vibrant street programs planned equitably?

(2) Were the plans implemented equitably?

Spatial equity here refers to whether the distribution of the planned and implemented programs prioritizes historically underserved neighborhoods, ensuring fair access to benefits such as enhanced safety, social infrastructure, and public amenities. We analyzed NYCDOT planned locations post-2021 for open streets, restaurants, plazas, street seats, bus stop shelters, newsstands, bike parking shelters, and automatic public toilets from the NYC Open Data portal. Of 54,061 citywide street segments, 9.4% were planned for at least one program, with Manhattan prioritized over other boroughs.

To assess implementation, we employed a local lightweight vision large language model (vLLM), Gemma 3, to analyze 4,720 street segments with paired Google Street View images (pre-2020 and 2024). The local vLLM compared image pairs to determine if planned programs were implemented, offering an efficient, consistent, and data-secure method for evaluating large datasets on standard computing resources. This approach minimizes reliance on external servers, enhancing privacy and scalability for urban planning analysis. Preliminary findings show low implementation rates for open streets, open restaurants and street seating programs, especially in the outer boroughs. In contrast, plazas, shelters, newsstands, bicycle parking, and restrooms had higher implementation rates citywide. The differences highlight the gap between the original intent of the vibrant street programs and their implementation.

This study highlights the critical need for ensuring equitable implementation of planned programs and improvements, including equity in stakeholder coordination efforts. Programs such as open streets require subsidies from local government, community-led activities, maintenance and enforcement through public works and law enforcement, and active participation by residents. These requirements impose a burden on implementation that is untenable in many neighborhoods, necessitating a rethink of open street program designs. Methodologically, this study presents a reliable, scalable method to monitor plan implementation. By quickly identifying implementation failures, planners can target interventions to reinstate programs effectively, advancing spatial equity and community vitality.

Vision Large Language Model