Who gets to use the street? Evaluate the utilization and inclusiveness using crowdsourced videos and vision-language models
Bibliographic record
Abstract
• The study proposes a framework for evaluating public street spaces. • We collect data from riders to build the largest known street dataset in Beijing. • We train a vision-language model to identify demographic groups and behaviors. • The study reveals uneven pedestrian patterns across age groups and street segments. Equitable access to public street spaces (PSS) represents a fundamental prerequisite for sustainable urban development, yet systematic evaluation frameworks remain inadequately developed. Existing studies suffer from demographic homogenization assumptions, and insufficient data resolution. The deficiency undermines the effective utilization and full inclusiveness of PSS, ultimately constraining street vitality, social cohesion, and progress toward sustainable development goals. We introduce a crowdsourced data collection method based on food delivery riders, significantly enhancing the detail and coverage of the dataset. Therefore, we collect over 11,000 hours of video and construct the “EgoCity Dataset”. Subsequently, we develop a powerful vision-language model (VLM) capable of identifying demographic groups and behaviors (F1-score: 0.9583), addressing critical gaps in existing computer vision applications for urban analysis. Finally, we propose a framework for evaluating PSS, which quantifies the actual level of urban vitality and spatial equity through two key indicators: utilization and inclusiveness. The main findings are: (1) At the individual level, pedestrian distribution shows significant demographic disparities, with adults dominating both volume and activity diversity. (2) At the street level, there is a spatial mismatch between pedestrian flow and residential population. (3) At the regional level, spatial advantages and resources are concentrated in select areas at the expense of broader equity. Our study proposes a low-cost data collection method, and is the first to integrate crowdsourced street video data with vision-language models for pedestrian group identification, as well as to introduce a novel framework for evaluating utilization and inclusiveness in public street spaces. This study offers practical tools for urban planning and advances the sustainable development goals.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.003 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".