Play the introduction video
Hover to compose a score from CV-detected pedestrian behavior
1/ Pedestrian tracking and classification
Walkers: speed ≥ 0.5 m/s
Lingerers: speed < 0.5 m/s for ≥ 5s
2/ Group detection
Within 1.9 meters for ≥ 2s and
move in same direction or all linger
3/ Identification of encounters
People arrive alone
Form a group ≥ 5s after entry
pedestrian speed: 1.23 → 1.41 m/s
% of single pedestrians: 79%
group encounters: 5.5% → 2%
upper: 2010s, lower: 1980s
Street Scores uses computer vision to analyze changes in pedestrian behavior over a 30-year period in four public spaces located in New York, Boston, and Philadelphia. Building on William Whyte’s observational work from 1980, where he manually recorded pedestrian behaviors, we employ computer vision and deep learning techniques to examine video footage from 1980 and 2010.
The project has first been presented at the 19th International Architecture Exhibition of La Biennale Architettura di Venezia, as Eyes on the Street.
Salazar-Miranda, A., Fan, Z., Baick, M. B., Hampton, K. N., Duarte, F., Loo, B. P. Y., Glaeser, E. L., & Ratti, C. (2025). Exploring the social life of urban spaces through AI. Proceedings of the National Academy of Sciences.
The material on this web site can be used freely in any publication
provided that
1. It is duly credited as a project by the MIT Senseable City Lab
2. A PDF copy of the publication is sent tosenseable-press@mit.edu
We envision this project as an ongoing dialogue between audience, art,
and city.
Download the Street Scores Toolkit
Izzi Waitz, Jingrong Zhang, Fábio Duarte
Music and Choreographed Dance
with
MIT Open Space Programming, supported by
Council for the Arts at MIT (CAMIT)
More coming soon
