Transcript Computer Vision Systems for the Blind and Visually Disabled.

Vision for the Blind.
Stat 19 SEM 2. 263057202. Talk 1.
Alan Yuille. [email protected]
UCLA. Dept. Statistics and Psychology.
www.stat.ucla/~yuille
Goal of the Course.
• How can technology and neuroscience
help the blind and disabled?
• (A) Artificial Intelligence systems.
• (B) Brain Implants. Neural Prostheses.
• Introduce and review the state of the art.
• Today, an overview.
Course Mechanisms.
• Maximum work load is 3 hours per week.
• Grading: Pass/Fail.
• Requirement. Don’t miss more than 2
lectures.
• Reading assignments. Short reports.
My Research Interests are:
• Designing computer vision systems
(artificial intelligence).
• Understanding how biological vision
systems work.
• I also work with a company that builds
prototype systems for the blind.
• Collaborate with the Smith-Kettlewell Eye
Research Institute (San Francisco).
What is Vision?
• Aristotle “to know what is where by looking”.
• Vision is information processing.
• Light rays that reach your eyes carry information
about the external world.
• Vision decodes the light rays and reconstructs
the world.
• Vision is a very complex task. Harder than
Mathematics? More on vision next lecture.
What would you miss most if you
went blind?
• Probably the ability to detect and read
text.
• Navigation is possible using a cane and/or
guide dog. Recognize people by their
voices. Hear oncoming traffic.
Experiment: Spend an hour with your eyes
closed.
What can blind people see if they
recover their sight?
• Mike May is one of the very few people
who attained his sight after being blind.
• (Successful businessman & skier).
• But how well can he use vision to get
information about the world?
• Infant’s vision systems develop in stages.
Some abilities develop at 1 month, others
at 3 months, etc.
Mike May
• Mike May can see light, but required extensive
•
•
•
•
•
training before he could distinguish a cube from
a sphere.
He cannot recognize faces (even his wife and
children).
It took two years before his vision helped his
skiing.
Handout. Interview with Mike May.
Mike Mays’ Journal:
www.sendergroup.com/mike.htm
Radio Interview. http://psy.ucsd.edu/~fine
How Big is the Problem?
• Legally Blind population of the US is 600900,000. (20/200 best correctly acuity).
• At least, 3,000,000 people have low
vision.
• These numbers are predicted to double in
the next ten years!
• Eyesight gets worse with age – and
people live longer.
Artificial Intelligence Solutions.
• Navigation: Global
Positioning Satelites
(GPS). (Loomis.
UCSB).
Computer Vision
• Digital Camera +
Portable Computer +
Speech Synthesizer.
Brain Implants.
• Artificial Retina:
• Connected to (i) the
•
•
retina, (ii) the optic
nerve, or (iii) the
cortex.
Dr. Dobelle.
Portugal.
Vision to Tactile.
• Dr. Bach y Rita. (Ex: Smith•
•
•
Kettlewell).
1970. Project an image onto
the back of a blind person.
Claim: could recognize
faces.
Few tactile receptors on
back. Better to use
fingertips, lips, or tongue.
Erik Weihenmayer. Climber.
Mt. Everest.
Neural Prostheses:
• Control by Thought.
• Movement Planning
•
Neurons.
Clint Eastwood film.
Firefox.
Summary.
• The course will describe these artificial
•
•
•
•
•
intelligence and biological techniques.
Can describe other topics: e.g. Cochlear
implants.
Mike Mays’s handout. Web journal:
www.sendergroup.com/mike.htm
Radio Interview: http://psy.ucsd.edu/~fine
Contact information: [email protected]