Five Principles for Choosing Research Problems in …funk/award_talk.pdfWhy this Topic? Possible...
59
Five Principles for Choosing Research Problems in Computer Graphics Thomas Funkhouser Princeton University
Transcript of Five Principles for Choosing Research Problems in …funk/award_talk.pdfWhy this Topic? Possible...
Five Principles for
Choosing Research Problems
in Computer Graphics
Thomas Funkhouser
Princeton University
Why this Topic?
Possible alternative topics I did this, then I did that, then …. zzz
What are the five biggest open problems … ???
My talk is somewhere in the middle Suggest five principles for choosing research problems
Explain how I’ve used these principles in my own choices
Five Principles
1. Consider non-traditional problems
2. Anticipate disruptive technologies
3. Think beyond algorithms
4. Work on real applications
5. Work with good collaborators
1. Consider Non-Traditional Problems
1. Consider Non-Traditional Problems
Computer GraphicsResearch Problems
1. Consider Non-Traditional Problems
Computer GraphicsResearch Problems
Previous Work
1. Consider Non-Traditional Problems
Computer GraphicsResearch Problems
Previous Work
ImproveExistingMethods
1. Consider Non-Traditional Problems
Computer GraphicsResearch Problems
Previous Work
ImproveExistingMethods
Parallel Radiosity
1. Consider Non-Traditional Problems
Computer GraphicsResearch Problems
Previous Work
ImproveExistingMethods
Develop New Approaches
to Existing Problems
1. Consider Non-Traditional Problems
Computer GraphicsResearch Problems
Previous Work
ImproveExistingMethods
Develop New Approaches
to Existing Problems
Modeling by Example
1. Consider Non-Traditional Problems
Computer GraphicsResearch Problems
Previous WorkDefine new problems
Develop New Approaches
to Existing Problems
ImproveExistingMethods
1. Consider Non-Traditional Problems
Computer GraphicsResearch Problems
Previous WorkDefine new problems
ImproveExistingMethods
Develop New Approaches
to Existing Problems
Shape Retrieval & Analysis
1. Consider Non-Traditional Problems
Computer GraphicsResearch Problems
Previous Work
Expand the field
ImproveExistingMethods
Define new problems
Develop New Approaches
to Existing Problems
1. Consider Non-Traditional Problems
Computer GraphicsResearch Problems
Previous Work
Expand the field
ImproveExistingMethods
Define new problems
Simulating Sound
Propagation
Develop New Approaches
to Existing Problems
Simulating Sound
Propagation
1. Consider Non-Traditional Problems
Computer GraphicsResearch Problems
Previous Work
Expand the field
ImproveExistingMethods
Define new problems
Develop New Approaches
to Existing Problems
1. Consider Non-Traditional Problems
Computer GraphicsResearch Problems
Previous Work
Expand the field
Define new problems
Develop New Approaches
to Existing Problems
ImproveExistingMethods
2. Anticipate Disruptive Technologies
2. Anticipate Disruptive Technologies
Time (in years)
Res
ea
rch
Pro
gre
ss
Clayton M. Christensen, 1995
2. Anticipate Disruptive Technologies
Time (in years)
Res
ea
rch
Pro
gre
ss
Clayton M. Christensen, 1995
2. Anticipate Disruptive Technologies
Time (in years)
Res
ea
rch
Pro
gre
ss
Fast GraphicsProcessors
Repositoriesof 3D Models
Clayton M. Christensen, 1995
Server-BasedNetworking
2. Anticipate Disruptive Technologies
Time (in years)
Res
ea
rch
Pro
gre
ss
Repositoriesof 3D Models
Clayton M. Christensen, 1995
Server-BasedNetworking
Fast GraphicsProcessors
2. Anticipate Disruptive Technologies
Time (in years)
Res
ea
rch
Pro
gre
ss
BuildingWalkthroughs
“Soda Hall Walkthrough”SGI PowerSeries with VGX Graphics64MB, 30K polygons/second~$200,0001992
Repositoriesof 3D Models
Clayton M. Christensen, 1995
Server-BasedNetworking
Fast GraphicsProcessors
2. Anticipate Disruptive Technologies
Time (in years)
Res
ea
rch
Pro
gre
ss
Fast GraphicsProcessors
BuildingWalkthroughs
Repositoriesof 3D Models
Clayton M. Christensen, 1995
Server-BasedNetworking
2. Anticipate Disruptive Technologies
Time (in years)
Res
ea
rch
Pro
gre
ss
Fast GraphicsProcessors
BuildingWalkthroughs
Repositoriesof 3D Models
Massively MultiplayerOnline Worlds
“RING”SGI Origin ServerHome: 28.8kb modem1994
Clayton M. Christensen, 1995
Server-BasedNetworking
2. Anticipate Disruptive Technologies
Time (in years)
Res
ea
rch
Pro
gre
ss
Fast GraphicsProcessors
BuildingWalkthroughs
Repositoriesof 3D Models
Massively MultiplayerOnline Worlds
Clayton M. Christensen, 1995
Server-BasedNetworking
Repositoriesof 3D Models
2. Anticipate Disruptive Technologies
Time (in years)
Res
ea
rch
Pro
gre
ss
Fast GraphicsProcessors
BuildingWalkthroughs
Repositoriesof 3D Models
Massively MultiplayerOnline Worlds
“Shape Distributions”133 models25 categories2001
Shape Retrieval& Analysis
Clayton M. Christensen, 1995
Server-BasedNetworking
Repositoriesof 3D Models
2. Anticipate Disruptive Technologies
Time (in years)
Res
ea
rch
Pro
gre
ss
Fast GraphicsProcessors
Repositoriesof 3D Models
Massively MultiplayerOnline Worlds
Shape Retrieval& Analysis
???
???
BuildingWalkthroughs
Clayton M. Christensen, 1995
Server-BasedNetworking
???
???
3. Think Beyond Algorithms
3. Think Beyond Algorithms
There are many ways to contribute to research: Develop algorithms
Design systems
Collect data sets
Develop benchmarks
Formulate problems
Prove theorems
Write surveys
etc.
3. Think Beyond Algorithms
There are many ways to contribute to research: Develop algorithms
Design systems
Collect data sets
Develop benchmarks
Formulate problems
Prove theorems
Write surveys
etc.
3. Think Beyond Algorithms
There are many ways to contribute to research: Develop algorithms
Design systems
Collect data sets
Develop benchmarks
Formulate problems
Prove theorems
Write surveys
etc.
www.cvpapers.com/datasets.html
3. Think Beyond Algorithms
Some of my projects focused on collecting data sets:
Schelling Points[Chen12]
Shape Retrieval[Shilane04]
Line Drawing[Cole08]
Mesh Segmentation[Chen09]
Sound Propagation[Tsingos02]
Shape Perception[Cole09]
3. Think Beyond Algorithms
Some projects focused on collecting data sets:
Surface Salience[Chen12]
Shape Retrieval[Shilane04]
Line Drawing[Cole08]
Mesh Segmentation[Chen09]
Sound Propagation[Tsingos2]
Shape Perception[Cole09]
3. Think Beyond Algorithms
Example: “Where Do People Draw Lines?”
3. Think Beyond Algorithms
Example: “Where Do People Draw Lines?”
Collect data while people
perform a task
3. Think Beyond Algorithms
Example: “Where Do People Draw Lines?”
Collect data while people
perform a task
Analyze data
3. Think Beyond Algorithms
Example: “Where Do People Draw Lines?”
Collect data while people
perform a task
Analyze data Evaluate existing algorithms
3. Think Beyond Algorithms
Example: “Where Do People Draw Lines?”
Collect data while people
perform a task
Analyze data Evaluate existing algorithms Learn to perform task
3. Think Beyond Algorithms
Example: “Where Do People Draw Lines?”
Analyze data Evaluate existing algorithms Learn to perform task
Collect data while people
perform a task
4. Work on Real Applications
4. Work on Real Applications
Core Computer Graphics
Technologies
Applications
4. Work on Real Applications
Core Computer Graphics
Cameras
Displays
Processors
Scanners
Sensors
Trackers
Printers
4. Work on Real Applications
Core Computer Graphics
Cameras
Displays
Processors
Scanners
Sensors
Trackers
Printers
Biology?
Psychology?Archaeology?
Paleontology?
Product Design? Education?
Neuroscience?
Civil Engineering?
Geology? Medicine?
4. Work on Real Applications
Core Computer Graphics
Cameras
Displays
Processors
Scanners
Sensors
Trackers
Printers
“The Computer Scientist as a Toolsmith”Frederick P. Brooks, Jr.
ACM Allen Newell Award PresentationSIGGRAPH 1994
Biology?
Psychology?Archaeology?
Paleontology?
Product Design? Education?
Neuroscience?
Civil Engineering?
Geology? Medicine?
4. Work on Real Applications
Some projects focused on applications:
Molecular Biology[Capra09] Archaeology
[Shin12]
Paleontology[Boyer11]
4. Work on Real Applications
Some projects focused on applications:
Molecular Biology[Capra09] Archaeology
[Shin12]
Paleontology[Boyer11]
Using partial shape matching to help find proteins with similar functions
DistinctiveRegions
Spherical HarmonicShape Descriptors
4. Work on Real Applications
Some projects focused on applications:
Molecular Biology[Capra09] Archaeology
[Shin12]
Paleontology[Boyer11]
Using a surface dissimilarity metric to help classify scans of fossils
4. Work on Real Applications
Some projects focused on applications:
Molecular Biology[Capra09] Archaeology
[Shin12]
Paleontology[Boyer11]
Using shape matching to help reconstruct fractured wall paintings
4. Work on Real Applications
Some projects focused on applications:
Molecular Biology[Capra09] Archaeology
[Shin12]
Paleontology[Boyer11]
5. Work With Good Collaborators
5. Work With Good Collaborators
Postdocs: Daniel Aliaga
Jian Sun
Sid Chaudhuri
Yaron Lipman
Ph.D. students: Aleksey Boyko
Aleksey Golovinskiy
Elena Sizikova
Fisher Yu
Maciej Halber
Michael Kazhdan
Patrick Min
Paul Calamia
Philip Shilane
Rudrajit Samanta
Tianqiang Liu
Vladimir Kim
Xiaobai Chen
Undergrad students: Abulhair Saparov
Addy Ngan
Alex Halderman
Alex Limpaecher
Angela Dai
Bill Pang
Carolyn Chen
Celeste Fowler
Hijung Shin
Jeehyung Lee
Joyce Chen
William Kiefer
Wilmot Li
5. Work With Good Collaborators
Diego Nehab
Emil Praun
Forrester Cole
Huiwen Chang
Jason Lawerence
Jingwan Lu
Joshua Podolak
Linjie Liu
Ohad Fried
Michael Burns
Sema Berkiten
Thiago Pereira
Wagner Correa
Xinyi Fan
Yiming Liu
Zhiyan Liu
Postdocs: Pierre Bénard David Bourgiuignon Martin Fuchs
Lee Markosian Tim Weyrich
Adam Finkelstein
SzymonRusinkiewicz
DavidDobkin
Ph.D. Students: Allison Klein
Benedict Brown
Chris DeCoro
Connelly Barnes
Corey Toler-Franklin
5. Work With Good Collaborators
Other Disciplines: Andreas Vlachopoulos
Ben Shedd
Christos Doumas
Doug Boyer
Gary Elko
Heather Barros
James West
Janet Thornton
Jukka Jernvall
Manish Singh
Mohan Sondhi
Otto J Anshus
Peter Svensson
Roman Laskowski
UC Berkeley: Carlo Séquin
Seth Teller
Bell Labs: Ingrid Carlbom
Alex Biliris
Gopal Pingali
Nicolas Tsingos
T.M. Murali
Stanford: Daniel Ritchie
Manolis Savva
Matthew Fisher
Pat Hanrahan
Qixing Huang
Other Institutions: Aaron Hertzmann
Ayellet Tal
Dan Goldman
David Jacobs
Doug DeCarlo
Eli Schechtman
Evangelos Kalogerakis
Hanspeter Pfister
Jim McCann
Leonidas Guibas
Niloy Mitra
Raif Rustamov
Stephen DiVerdi
Wilmot Li
Wojciech Matusik
5. Work With Good Collaborators
Other Disciplines: Andreas Vlachopoulos
Ben Shedd
Christos Doumas
Doug Boyer
Gary Elko
Heather Barros
James West
Janet Thornton
Jukka Jernvall
Manish Singh
Mohan Sondhi
Otto J Anshus
Peter Svensson
Roman Laskowski
UC Berkeley:Carlo Séquin
Seth Teller
Bell Labs: Alex Biliris
Ingrid Carlbom
Gopal Pingali
Nicolas Tsingos
T.M. Murali
Stanford: Daniel Ritchie
Manolis Savva
Matthew Fisher
Pat Hanrahan
Qixing Huang
Other Institutions: Aaron Hertzmann
Ayellet Tal
Dan Goldman
David Jacobs
Doug DeCarlo
Eli Schechtman
Evangelos Kalogerakis
Hanspeter Pfister
Jim McCann
Leonidas Guibas
Niloy Mitra
Raif Rustamov
Stephen DiVerdi
Wilmot Li
Wojciech Matusik
5. Work With Good Collaborators
Other Disciplines: Andreas Vlachopoulos
Ben Shedd
Christos Doumas
Doug Boyer
Gary Elko
Heather Barros
James West
Janet Thornton
Jukka Jernvall
Manish Singh
Mohan Sondhi
Otto J Anshus
Peter Svensson
Roman Laskowski
UC Berkeley: Carlo Séquin
Seth Teller
Bell Labs: Alex Biliris
Ingrid Carlbom
Gopal Pingali
Michael Potmesil
Nicolas Tsingos
Stanford: Daniel Ritchie
Manolis Savva
Matthew Fisher
Pat Hanrahan
Qixing Huang
Other Institutions: Aaron Hertzmann
Ayellet Tal
Dan Goldman
David Jacobs
Doug DeCarlo
Eli Schechtman
Evangelos Kalogerakis
Hanspeter Pfister
Jim McCann
Leonidas Guibas
Niloy Mitra
Raif Rustamov
Stephen DiVerdi
Wilmot Li
Wojciech Matusik
5. Work With Good Collaborators
Other Disciplines: Andreas Vlachopoulos
Ben Shedd
Christos Doumas
Doug Boyer
Gary Elko
Heather Barros
James West
Janet Thornton
Jukka Jernvall
Manish Singh
Mohan Sondhi
Otto J Anshus
Peter Svensson
Roman Laskowski
UC Berkeley: Carlo Séquin
Seth Teller
Bell Labs: Ingrid Carlbom
Alex Biliris
Gopal Pingali
Nicolas Tsingos
T.M. Murali
Stanford: Daniel Ritchie
Manolis Savva
Matthew Fisher
Pat Hanrahan
Qixing Huang
Other Institutions: Aaron Hertzmann
Ayellet Tal
Dan Goldman
David Jacobs
Doug DeCarlo
Eli Schechtman
Evangelos Kalogerakis
Hanspeter Pfister
Jim McCann
Leonidas Guibas
Niloy Mitra
Raif Rustamov
Stephen DiVerdi
Wilmot Li
Wojciech Matusik
5. Work With Good Collaborators
Other Disciplines: Andreas Vlachopoulos
Ben Shedd
Christos Doumas
Doug Boyer
Gary Elko
Heather Barros
James West
Janet Thornton
Jukka Jernvall
Manish Singh
Mohan Sondhi
Otto J Anshus
Peter Svensson
Roman Laskowski
UC Berkeley: Carlo Séquin
Seth Teller
Bell Labs: Ingrid Carlbom
Alex Biliris
Gopal Pingali
Nicolas Tsingos
T.M. Murali
Stanford: Daniel Ritchie
Manolis Savva
Matthew Fisher
Pat Hanrahan
Qixing Huang
Other Institutions: Aaron Hertzmann
Ayellet Tal
Dan Goldman
David Jacobs
Doug DeCarlo
Eli Schechtman
Evangelos Kalogerakis
Hanspeter Pfister
Jim McCann
Leonidas Guibas
Niloy Mitra
Raif Rustamov
Stephen DiVerdi
Wilmot Li
Wojciech Matusik
5. Work With Good Collaborators
Family:
5. Work With Good Collaborators
Thank You!
