Accepted paper ·
Alina Hagen's street-view accessibility paper accepted to WACV 2027
"Do VLMs See What Sensors Feel?" — an expert-guided design that scores wheelchair accessibility from Google Street View and checks it against GPS-derived wheelchair dwell behavior — will be presented at IEEE/CVF WACV 2027.
Do VLMs See What Sensors Feel? A Scalable Expert-Guided Design for Wheelchair Accessibility Assessment from Street View — by Dongdong Wang, lab PhD student Alina Hagen, Isabelle Gatmaitan, Hao Zhou, Yiwen Dong, Shabboo Valipoor, Vivian W.H. Wong and Lingyao Li — has been accepted to the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV 2027), the lab’s second WACV 2027 paper alongside DriveMotion.
The paper asks whether vision-language models can identify accessibility barriers from Google Street View imagery. An expert-guided, retrieval-augmented framework combines street-view images, ADA-informed guidance and expert-derived rubrics, and is evaluated on a campus-scale dataset at the University of Florida that links 407 street-view locations with GPS-derived wheelchair dwell behavior as a mobility-friction signal. VLM ratings are negatively correlated and distributionally similar with dwell time — partial but consistent alignment with a behavioral proxy — while subtle surface conditions, transient obstructions and viewpoint-dependent barriers remain hard to see.