Personal Object Recognizers for Blind People Receive Honorable Mention at CHI 2017[/caption]
Robotics Institute faculty Kris Kitani and Chieko Asakawa in collaboration with Human-Computer Interaction Institute postdoctoral fellow Hernisa Kacorri and faculty Jeffrey Bigham received a Best Paper Honorable Mention for their project helping blind people identify objects at the ACM CHI Conference on Human Factors in Computing Systems.
Their research explores personal object recognizers, an approach that allows blind people train an object recognition algorithm to differentiate between everyday objects specific to the user by providing a few snapshot of the objects in the user’s environment. By doing so, they aim to provide a solution for a very practical need for blind people – instance level object recognition.
Despite challenges in photo taking, blind participants were found to train and adapt a deep learning system to recognize everyday objects at accuracies comparable to sighted people.