MassDOT tested an AI framework developed with UMass Lowell that analyzes aerial photographs to automatically detect and classify crosswalks across the entire state of Massachusetts, identifying approximately 88,000 crosswalks in 2021 imagery.

The project used annotated aerial images downloaded from MassGIS to train an AI computer vision model, which was then applied to statewide images from both 2019 and 2021. The model classified crosswalks by type (continental/zebra-style, parallel lines, or solid) and location category (intersection, midblock, or driveway). The resulting model and post-processing scripts were made available to MassDOT for further analyses. Results were expected to inform maintenance and safety planning. As of the 2025 Q1 newsletter, MassDOT listed 'AI for Sidewalk Detection' as an active research initiative under a University of Massachusetts Lowell agreement.