Uploaded October 2020 | Updated September 2026, 1 hour ago
Inaccessible entrances to buildings pose a challenge for wheelchair users. A single step can mean that wheelchair users cannot enter a building, whether to attend an event, eat at a restaurant, or visit a friend. Can machine learning help by educating business owners and forewarning wheelchair users? In this session, I'll present findings from my project using machine vision to identify features indicating (in)accessibility, including steps, stairs, thresholds, uneven pavement, ramps, and automatic doors. The model, built using Google's Cloud AutoML Vision Object Detection, used thousands of hand-labeled images of building entrances. I'll share how attendees can build similar models for themselves, and offer tips on interpreting results.
Inaccessible entrances to buildings pose a challenge for wheelchair users. A single step can mean that wheelchair users cannot enter a building, whether to attend an event, eat at a restaurant, or visit a friend. Can machine learning help by educating business owners and forewarning wheelchair users? In this session, I'll present findings from my project using machine vision to identify features indicating (in)accessibility, including steps, stairs, thresholds, uneven pavement, ramps, and automatic doors. The model, built using Google's Cloud AutoML Vision Object Detection, used thousands of hand-labeled images of building entrances. I'll share how attendees can build similar models for themselves, and offer tips on interpreting results.
