2. Computer Vision at Microsoft
• Photo editing (stitching, PhotoFuse, GrabCut)
• Photo Tourism → Photosynth
• Maps: photogrammetry, stitching
• Mobile recognition: product search, OCR
• Mobile (computational) photography
• Kinect
• Medical image analysis (Amalga)
3. Tech transfer at Microsoft
“Classic” 3-stage push model:
1. Research papers (stitching, PhotoMontage, Grab
Cut, Photo Tourism)
2. Prototype or incubation
(ICE, GroupShot, Photosynth, Lincoln)
3. Product
But also works other way (product pull):
– Kinect (secret project, hand-selected researchers)
– Amalga medical image analysis
4. Microsoft - Academia
• Microsoft Research Connections
• Microsoft Research Faculty Fellows
• Microsoft Research PhD Fellows
• Internships
• Faculty Summit
5. Improving relations (I)
• More accessible tutorials / teaching materials for
non-researchers:
– tutorials at conferences (will people attend?)
– on-line courses, exercises
• Better libraries:
– standard libraries (like OpenGL)
– free, non-commercial, commercial licenses
• Researcher training:
– efficient algorithms & coding (software engr.)
– scenario-driven research
– technical communications
6. Improving relations (II)
• More information flow industry → academia
– panels at conferences
– David’s list of computer vision companies
• encourage groups to list of areas and open problems,
e.g., http://www.disneyresearch.com/research/index.htm
• Funding models and IP
– tough one: lots of models, contracts vs. open gifts
– fellowships (few), internships (many)
– IP tricky both ways