Project Sidewalk: Designing the future of public sidewalks
A platform to help urban planners detect and manage sidewalk obstructions using machine learning and computer vision.
Visit websiteThe Problem
For people with disabilities, navigating the world can be a challenge. One of the biggest obstacles is the lack of accessibility on public sidewalks. This is especially true in cities where there are many obstructions such as construction sites, parked cars, and overgrown vegetation. Thus, this project aims to help urban planners and city officials better understand the accessibility of their sidewalks so that they can make improvements.
The Labelling System
To streamline the process of labelling sidewalk obstructions, we created a web application that allows users to label obstructions using satellite imagery. The application is built using jQuery and Scala and is deployed using Docker and Linux. This helps build models that will soon automate and classify obstructions quicker and provide urban planners with more information about the accessibility of their sidewalks.
Validating Labels
To ensure the quality of the labels, we created a validation system that allows users to validate labels made by other users. This helps ensure that the labels are accurate and that the models will be trained correctly. We can use this to generate true positives for our models.
Machine Learning
We can then use this data that is labelled and validated using the Project Sidewalk application to train machine learning models to detect and classify sidewalk obstructions.
Encouraging Accessibility
Over 93.8% of target areas have been labelled and validated by users, covering over 1,518.6 total miles covered. This data is crucial and will ultimately help people who have difficulty navigating the world.
Project outcomes
Ultimately, Project Sidewalk aims to create a path for everyone. We hope that this project will help urban planners and city officials better understand the accessibility of their sidewalks so that they can make improvements. We are currently in five cities across the world, and rapidly expanding to more cities around the world with over 2 million dollars in grant funding.