Crowdsourced labeled image data, such as from University of Washington's Project Sidewalk, could support assistive AR development. For instance, computer vision and ML models being trained for autodetection might be leveraged to augment real-time pedestrian accessibility aids, and next generation AR glasses could provide a means for individuals to contribute to labeling and validation in real-time. Such forms of technology could expand the ways accessibility information is collected and provide greater immediacy to labeling and validation efforts.