VizWiz answers questions that people who are blind or visually impaired have about the things around them. Users take a photo, ask a question, and receive answers back quickly from people on the web (the crowd). VizWiz leverages color identification, text recognition, and recognition of objects seen before, but automatic methods are limited to a small subset of questions in practice. The bulk of remaining questions are answered by humans, e.g. “is there a rash on my baby’s head,” “what number is on this credit card,” serving as challenges for computer vision. The VizWiz living laboratory illustrates the utility of deploying working crowd-powered systems to understand target domains via deployable Wizard-of-Oz. On-going work is using VizWiz to explore how people may volunteer their friends for microtasks they care about on social media.
Link to Project
Learning Sciences and Technologies