This project will develop an automated system to detect and characterize intersection markings and assess their condition using currently-available roadway geographic information system (GIS) data and aerial images. Work in Stage 1 will focus on collecting training data and developing core modules for data acquisition and computer vision. A data acquisition module will be developed to automatically retrieve intersection locations from roadway GIS data and capture corresponding aerial images. The input datasets are mostly available from agencies’ public databases or open sources such as Google Maps and OpenStreetMap. Marking images will be synthesized from different environment settings, and the synthesized data will be used to pre-train computer vision models for marking detection and characterization. A multi-task deep learning model will be built that would embed conventional neural network for marking detection, characterization, and assessment of marking degradation. Work in Stage 2 will involve prototype development and testing and demonstration. The prototype will be built by integrating the modules developed in Stage 1, and a web-based graphical user interface (GUI) will be developed for the system. The system will be initially tested in the laboratory setting using data from a small set of target areas followed by testing with a large-scale road network in Virginia.