Using Computer Vision to Support Accessibility for People with Visual Impairments
As computer vision grows in power and reliability it is increasingly used in applications to support accessibility for people with visual impairments, such as smartphone apps that perform OCR, product recognition and scene recognition. Drawing on my experience using computer vision to support accessibility, I will describe how this application domain must be informed by an understanding of the tools and techniques already available to people with visual impairments, and by an appreciation of which needs are not yet met satisfactorily. The presentation will also include a discussion of the limitations of computer vision, together with ways of harnessing additional techniques and technologies to overcome these limitations. Finally, I will describe possible opportunities for students to get involved in applying computer vision and other technologies to support accessibility.
James M. Coughlan received his B.A. in physics at Harvard University in 1990 and completed his Ph.D. in physics there in 1998. He is currently a Senior Scientist at The Smith-Kettlewell Eye Research Institute in San Francisco, California. His main research focus is the use of computer vision and sensor technologies to facilitate greater accessibility of the physical environment for blind and visually impaired persons. Current and past accessibility projects include the development of systems to provide audio-haptic access to physical objects such as documents and 3D models, the ability to find and read signs and other textual information, and navigation assistance indoors and at traffic intersections. He is also interested in probabilistic methods in computer and human vision, particularly the use of graphical models and belief propagation for visual search, segmentation and 3D stereo.