MIRACLE: Motor ImpaRiment AcCessibiLity issuEs

Abstract

Mobile app accessibility is the concept that a given application should function equally well for all people, regardless of potential impairments. The Web Content Accessibility Guidelines, which all accessibility standards are based on, provide detailed instructions on how app owners can make platforms accessible to users with varying disabilities. After analyzing the prevalence of accessibility issues in over 100 Android apps we found that almost all apps are riddled with accessibility issues, hindering their use by individuals with disabilities. Through this project, we aim to build a motor impairment-focused developer tool, called MIRACLE, that is able to analyze an app screenshot and metadata and output a full accessibility report describing design issues that could affect motor-impaired users. Our tool is able to identify violations of five common guidelines for building apps that are accessible to both motor-impaired users and adhere to universal design principles. After collecting data consisting of mobile app screenshots and corresponding metadata, we created a dataset with violations of all five guidelines. We have finished the implementation of our tool which uses deep learning and computer vision techniques to identify violations of accessibility guidelines. We plan to evaluate the overall usability of Miracle by measuring how effectively it can detect accessibility issues in popular open-source apps.