Tile2Net
An end-to-end tool for automated mapping of pedestrian infrastructure from aerial imagery.
Tile2Net
An end-to-end tool for automated mapping of pedestrian infrastructure from aerial imagery.
CitySurfaces
Open-source active learning-based framework for sidewalk materials segmentation from street level imagery, reducing annotation effort through uncertainty sampling and transparent failure diagnostics.
CitySurfaces
Open-source active learning-based framework for sidewalk materials segmentation from street level imagery, reducing annotation effort through uncertainty sampling and transparent failure diagnostics.
Crosswise
A context aware similarity retrieval platform that combines multimodal data, computer vision, spatial analysis, and graph based learning to find comparable intersections across jurisdictions and examine how design and surrounding urban context relate to safety, accessibility, and equity.
Crosswise
A context aware similarity retrieval platform that combines multimodal data, computer vision, spatial analysis, and graph based learning to find comparable intersections across jurisdictions and examine how design and surrounding urban context relate to safety, accessibility, and equity.
StreetTransformer
Uses historical imagery, geospatial data, and pedestrian infrastructure signals to distinguish meaningful changes in sidewalks, crossings, street geometry, and the public right of way.
StreetTransformer
Uses historical imagery, geospatial data, and pedestrian infrastructure signals to distinguish meaningful changes in sidewalks, crossings, street geometry, and the public right of way.
Humanitarian Infrastructure Mapping for Food Security
Fine scale map of Zambia's paved and unpaved roads, and building footprints to support food aid targeting using custom trained computer vision model.
Humanitarian Infrastructure Mapping for Food Security
Fine scale map of Zambia's paved and unpaved roads, and building footprints to support food aid targeting using custom trained computer vision model.
GeoJupyter
Open source tools that connect spatial data discovery, computation, mapping, and publication into one reproducible Jupyter workflow.
GeoJupyter
Open source tools that connect spatial data discovery, computation, mapping, and publication into one reproducible Jupyter workflow.
GeoJupyter Trail
A guided, teachable workflow that follows a real geospatial project from raw data to reproducible publication.
GeoJupyter Trail
A guided, teachable workflow that follows a real geospatial project from raw data to reproducible publication.
ELSA
Evaluates open-vocabulary and vision-language models for spatially grounded urban analysis of social activities and street conditions. Examines when model category structures are stable enough to support planning research and when they impose inappropriate abstractions on urban scenes.
ELSA
Evaluates open-vocabulary and vision-language models for spatially grounded urban analysis of social activities and street conditions. Examines when model category structures are stable enough to support planning research and when they impose inappropriate abstractions on urban scenes.
Complete Trip Accessibility
Reframes accessibility measurement from proximity to route-level feasibility for people with different mobility needs. Integrates pedestrian network topology, transit connectivity, surface conditions, and physical barriers into a unified model of first-and-last-mile navigability.
Complete Trip Accessibility
Reframes accessibility measurement from proximity to route-level feasibility for people with different mobility needs. Integrates pedestrian network topology, transit connectivity, surface conditions, and physical barriers into a unified model of first-and-last-mile navigability.
SideSeeing
Multimodal dataset combining video, inertial measurement, and GPS sensing for sidewalk condition assessment. Captures fine-grained pavement quality signals that complement imagery-based approaches and field audits, supporting reproducible accessibility research across urban environments.
SideSeeing
Multimodal dataset combining video, inertial measurement, and GPS sensing for sidewalk condition assessment. Captures fine-grained pavement quality signals that complement imagery-based approaches and field audits, supporting reproducible accessibility research across urban environments.
StreetAware
Synchronized audio, video, and LiDAR sensing at intersections supporting research on pedestrian safety and accessibility.
StreetAware
Synchronized audio, video, and LiDAR sensing at intersections supporting research on pedestrian safety and accessibility.
Heat Exposure Analysis
Sidewalk-level heat risk assessment integrating pedestrian route-choice modeling with urban microclimate simulation. Maps where heat exposure concentrates along walking routes at finer spatial resolution than conventional studies, identifying heat-vulnerable corridors for planning and public health response.
Heat Exposure Analysis
Sidewalk-level heat risk assessment integrating pedestrian route-choice modeling with urban microclimate simulation. Maps where heat exposure concentrates along walking routes at finer spatial resolution than conventional studies, identifying heat-vulnerable corridors for planning and public health response.
Proximity
Network analysis and pathfinding engine for modeling active transport accessibility at the route level, serving as the front end of the Profilo platform. Supports pedestrian and cycling network analysis with accessibility metrics calibrated to real mobility constraints.
Proximity
Network analysis and pathfinding engine for modeling active transport accessibility at the route level, serving as the front end of the Profilo platform. Supports pedestrian and cycling network analysis with accessibility metrics calibrated to real mobility constraints.
The Illusion of Stable Features
Methodological examination of whether latent representations in geospatial deep learning retain meaning across models, places, and training conditions. Investigates spurious correlation and proposes validation standards for AI-derived urban evidence that go beyond benchmark performance.
The Illusion of Stable Features
Methodological examination of whether latent representations in geospatial deep learning retain meaning across models, places, and training conditions. Investigates spurious correlation and proposes validation standards for AI-derived urban evidence that go beyond benchmark performance.
Sidewalk Ballet
Large-scale analysis of social interaction patterns on urban sidewalks, drawing on Jane Jacobs's concept of sidewalk life as self-organizing choreography. Uses computer vision and spatial methods to study how people share and navigate public space at scale. A collaboration between MIT and HPI, funded by the Design for Sustainability Initiative at MIT.
Sidewalk Ballet
Large-scale analysis of social interaction patterns on urban sidewalks, drawing on Jane Jacobs's concept of sidewalk life as self-organizing choreography. Uses computer vision and spatial methods to study how people share and navigate public space at scale. A collaboration between MIT and HPI, funded by the Design for Sustainability Initiative at MIT.
