A Demonstration of RASSAR: Room Accessibility and Safety Scanning in Augmented Reality
Xia Su, Kaiming Cheng, Han Zhang, Jaewook Lee, Wyatt Olson, Jon E. Froehlich · 2023 · Proceedings of the 25th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS) · doi:10.1145/3597638.3614504
Summary
This demo paper introduces RASSAR (Room Accessibility and Safety Scanning in Augmented Reality), a mobile AR application that semi-automatically identifies, localizes, and visualizes indoor accessibility and safety issues using iPhone LiDAR sensors and real-time computer vision. The system addresses a critical problem: in the United States, 90% of housing units have accessibility issues such as entrance steps and narrow doorways that make them unsafe or inaccessible to people with mobility disabilities. While paper checklists like the Home Safety Self-Assessment Tool (HSSAT) exist for auditing indoor spaces, they are time-consuming and require manual measurement. RASSAR automates this process by having users slowly scan a room with their phone while the app constructs a real-time 3D parametric model using Apple's RoomPlan API and runs a custom-trained YOLOv5 object detection model. The system detects 20 types of accessibility and safety issues across four categories: inaccessible object dimensions (bed height, table height, counter height, door width, opening width), inaccessible object positions (cabinet, sink, knob, door handle, light switch, outlet, grab bar heights), presence of risky items (rugs, scissors, knives, medication), and lack of assistive devices (grab bars near toilet/tub, fire alarms). The design was informed by formative interviews with 18 participants from five stakeholder groups: wheelchair users, blind and low vision people, families with young children, older adults/caregivers, and occupational therapists. Users can customize scans by selecting target communities, with different issues flagged for different groups.
Key findings
Technical evaluation across eight home spaces (24 total scans) showed promising results: average accuracy of 0.71, precision of 0.86, recall of 0.81, and F1 score of 0.83, with substantial inter-scan consistency (Fleiss' Kappa of 0.76). Critically, RASSAR scanning averaged 106 seconds per room compared to approximately 10 minutes for manual auditing — a roughly 6x speed improvement. In the formative study, 16 of 18 participants were favorable toward the prototype, particularly valuing measurement and documentation features, the ability to prepare homes for visitors with accessibility needs, and customizable issue detection. The system provides real-time AR visualization of detected issues with pop-up icons, detailed information panels with recommended solutions, and a post-scan interactive 3D room reconstruction with detected issues highlighted in red. Results can be exported as JSON for further processing. The authors acknowledge important limitations: the scanning process itself requires holding a phone upright, seeing the space, and interpreting visual results, making it currently inaccessible to blind/low-vision users and some people with motor disabilities.
Relevance
RASSAR represents an innovative application of consumer AR technology (LiDAR-equipped smartphones) to indoor accessibility assessment, a domain traditionally reliant on manual professional audits or self-assessment checklists. The practical use cases are compelling: vacation rental hosts verifying accessibility claims, caregivers preparing homes for people with new disabilities, families childproofing spaces, and people with disabilities themselves documenting issues for renovation planning or accommodation requests. The customizable rubric approach — where different accessibility communities see different relevant issues — is a thoughtful design choice that avoids one-size-fits-all accessibility assessment. For the broader accessibility field, this work demonstrates how emerging mobile sensing technologies can democratize accessibility auditing, potentially moving it from a specialized professional activity to something any smartphone user can perform. The honest acknowledgment that the tool itself has accessibility barriers (requiring vision and dexterity to operate) points to important future work in making accessibility tools themselves accessible.
Tags: augmented reality · computer vision · indoor accessibility · object detection · LiDAR · home safety · accessibility auditing · mobile application
Standards referenced: ADA Design Standards · Fair Housing Act