Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

[SIGGRAPH ASIA 2011 Course]How to write a siggraph paper


Published on

I found this slide on the forum. Thx for the guy that wrote most of the content down for us to review. Hope everyone can learn and think a lot from it!

[SIGGRAPH ASIA 2011 Course]How to write a siggraph paper

  1. 1. How to write a SIGGRAPH paperReconstructed from SIGGRAPH Asia 2011 Course
  2. 2. Writing a paper is likean interview Expose it right Pick the right problem Write a Execute paper it right well
  3. 3. SIGGRAPH• What SIGGRAPH wants • What we have• Right problem • Do what you love• Novel idea • Do the best• Solid algorithms • Randomness• Good apps & results • Objectivity• Clear writing • Discipline• Engaging talk • Practice
  4. 4. What SIGGRAPH wants• Right problem • What people care right now• Novel ideas • Do not do incremental stuff • Think big, be creative (not easy)• Solid algorithms • aka technical contribution • Knowledge is power; study hard, be evil
  5. 5. • Good applications • Give people what they need• Good results/effects • Work hard (in demos)• Good writing • You can’t sell what you can’t tell • It is all about bullshit
  6. 6. Right problem• Do what people care• Do what people don’t hate • Many interesting/cute problems in graphics • Pick what you love• Don’t be afraid of difficult problems • Less competition • Be brave
  7. 7. Novel ideas• Don’t do incremental stuff • You are wasting your time• Aim for the best• Read as many as possible(SIGGRAPH, geometry, texture, rendering…)• Results/demos are important! Technically awesome. • Something the state-of-the-art cannot do • Comparison are often unavoidable
  8. 8. Solid algorithms• Practice • Most algorithms are modified from others • Read a lot of papers• Uncertainty • Don’t expect to get everything right on 1st try • Expect failures
  9. 9. Good applications• Be broad • Do what you love
  10. 10. Clear writing• Practice • Blog, love letters, etc• Discipline • Good writing takes time; don’t wait until the end• Objectivity • Have others read your drafts • It is hard to see what you know while others don’t
  11. 11. Mentality• Mentality more important than talent • Do what you love • Aim for the best• Train your mentality • Practice • Have fun
  12. 12. Refine the problem & convergeProblem Refine Last Month TimeRefine
  13. 13. Research Front• Age to Contribute.• Outsiders are more likely to come up with new ideas.• Naï students help! ve
  15. 15. SHIFT• From areas you know into other areas you don’t know.• Examples: • Detailed deformation  Image warping • Captcha  Image emerging
  16. 16. Adapt• Adopt from other field into your expertise• Examples: • L1 median  Image enhancing, filtering and surface reconstruction • Machine learning  3D analysis • Mean-value coordinate  Image editing, cloning
  17. 17. Expand• Expansion of an existing area• Examples: • Seam carving for image  For video • 2D vector texture  3D vector volume texture
  18. 18. Trends, opportunities• Identify new phenomena recently appeared, and employ them.• Examples: • Photo explosion on web  Scene completing using web searching, ``Sketch2photo”, photo tourism • Popularity of Wii  Kinect
  19. 19. Identify successfultechniques and apply• Internet  Image completion  Surface completion• Computational photography  Kinect• High-performance Computing  Mechanical Turk
  20. 20. Identify / Problems• Ray Tracing• Radiosity• Image Retargeting• Mesh Compression• Etc…
  21. 21. Retro• Pick a problem that was only hot 20 years ago and revisit it.• Example: • Occlusion Culling  City Visualization
  22. 22. Invent a New Problem
  23. 23. Reversea Known Problem• Synthesis large texture from small texture  Abstract from large texture to small texture (Inversed texture synthesis!)• Detect an object in the image  Hide classified information in the image• Image Colorization  Image Grayization
  24. 24. Find an UnexploredProblem• Depixelizing Pixel Art• Face Beautification
  25. 25. And Expose it Right!• Make it sound interesting• Surprise!!!• A general message
  26. 26. Terminologyis important• Final words • Make your readers’ life as easy as you can • Main thing: aim at innovative, impactful research work • Don’t despair!
  27. 27. Avoid the “delta”impression• Chrystal clear expression• Intuition is always helpful• Try to find the simplest understandable explanation for your math
  28. 28. Practical Tips forMath haircut• Clear and neat notation• Always define all symbols• Give equation numbers• Assume nothing, explain everything
  29. 29. How to get awaywith it?• Make sure you speak the language of the community• Do your homework-learn previous work• Use common terminology
  30. 30. Writing: get helpand learn from experienced people• Contribution/page radio • Single idea: 4 pages • 1.5 ideas: 6 pages • 2 ideas: >8 pages• Schedule your projects as early as possible• The best of luck && Take care of your health~~
  31. 31. If you are not highly experienced, you’d group with other 2~6people.
  32. 32. The multi-touch screen in your hands was invented in 70’s, as afailed product.
  33. 33. Live as BainingBaining Guo:• 1986: firstly heard CG• 1988: live with Eric Haines• 1989: met Dani Lischinski• 1990: attended D.Greenberg’s Graphics course• 1996: firstly attend SIGGRAPH conference• 1998: first SIGGRAPH paper• 1999 ~ now: more SIGGRAPH papers, • And much more rejected ones.
  34. 34. Our mission• Advance in each field we do research• Transfer new technologies into products• Ensure your lab a future!
  35. 35. Faster, higher,stronger! You only live once, so have some serious fun.
  36. 36. Kill, Kill,KILL!!! KILL, Kill!!!• Good ideas grow from killing bad ideas• Kill false or mediocre projects• Be ruthless• Focus on one high-quality work• Walk the fine line of greatness and stupidity• “Fail fast” you should! • If you feel trapped in your idea, usually it would be a bad idea.
  37. 37. You’d always: Collaborate andcontact with Product Developersand Users
  38. 38. WRITING IT UP
  39. 39. As you start• Break writer’s block, start with the body• From text to structures: • Relentlessly focus on what you’ve done and never try to impress everyone • Structure, structure, structure… • Be concise: “Appendix test”! • One section for one person if you are working as a group (hope you’re not writing alone!)
  40. 40. PerfectionismQuestion yourself:• Have you provided references or justification for whatever you stated?• For things difficult to evaluate mathematically, have you provided a user study?• Does your method have a lot of practical applications? Are they surprisingly fresh or just stereotyped ones?• Can’t your results be more pleasing?
  41. 41. “Introduction”• Where passion met facts• So be passionate!
  42. 42. “Result”• Keep your promise!!!
  43. 43. Other RULES• Don’t copy conclusion from abstract. • Here you’d have some more deeper view!• Never praise your own work! • Not “we present an elegant algorithm…”• Don’t (intentionally or unintentionally) hide problems! • Realize the problem and try to fix that (either in your paper or in your future work), not elude from that.
  44. 44. What if Rejected?• Never forget the long term review • Building a career is a long process. • Your physical and mental health come first. • Sustained good performance comes next. • Always act professionally. • Drastic local events are not big deals in long term.
  45. 45. What if Rejected?• So have a good rest!• Don’t bet your career on a SIGGRAPH paper!• Inspired by the reviews? • Resubmit.• Mature work? • Find your champions and get the work out ASAP.• Again anchor your decision on analysis of your work
  46. 46. Live with rejections• Do what you love, so that you won’t mind reject• Treat rejections as normal and routine• Life is not as fun without failures (I mean it)
  47. 47. Learn from yourrejections• You CAN learn from rejections• Learn to listen to your reviews, and filter out their outliers.• Learn to listen to your friends.
  48. 48. The review Form• Did the reviewer understand what the paper is about?• Contribution scope: How important is your work? Is widely applicable? Is there abundant analysis?• Contribution magnitude: amount of novelty, originality.• If the paper is poorly written, it always get rejection no matter how good the idea is.• It’s your responsibility to ensure the reviewers understand your paper, make their job easier!• Be kind/fair, avoid insulting previous methods!• Completeness, mention important implementation details, constant values.• Make sure you demonstrate/discuss any drawbacks or limitations
  49. 49. The importance ofthe Introduction cannot be more estimated• “Uneducated guess”: in over 90% the reviewers will have made up his mind while reading Introduction.• Goals: • What the paper is about? • What problems does it address? • Why should the reader care?
  50. 50. Convince• Your problem should be important. It has not been solved enough.• Apparently you have a novel solution.
  51. 51. Tips I• Demonstrate the problem solving• Show the shortcomings of existing methods• Visual aids to help explain• Demonstrate the quality• The reviewers should understand just from figures & captions.
  52. 52. Tips II• Always keeping promises • Never over/understate • Make a balance between your and previous work • Be through, be fair, and support your claim about their shortcomings, never write a laundry list for them• In your results, you’d point out benefit. Don’t assume the reviewers understand• In your conclusions, re-iterate limitations for future work and summarize what you’ve achieved.
  53. 53. Start trying when you are 22-25 years oldFailure with first couple of triesYour first  happiest day of your life!Afterwards, life settling into a routine …ROAD TO 1ST SIG PAPER
  54. 54. Like your first love• Idealistic: beautiful, even sexy, pretty name …• The hardest to get  few “love at first sight”• Devote your passion and patience
  55. 55. Do not sue me …• Be a little bit careful if you are married • It helps to be single …• or Chinese: new year always after deadline!
  56. 56. Making the first …• The right people• The right mindset• The right problem
  57. 57. People: mentor• Learn from and work with the experts• They know • What is a SIGGRAPH-able idea • How to make a SIGGRAPH paper • How to do that in an industrial lab • How to write a SIGGRAPH paper • …
  58. 58. People: student• One who can finish • Smart and hard-working • No genius  needs execution• One who can pay attention to details• One who has the sense of aesthetics• One One who write,notleast the technical week! who can does at wait until the last parts
  59. 59. People: colleague• Not an exact match with your expertise• Those who complement you • Machine learning, statistics, optimization • Differential geometry• Those who brings you surprising problems • Architects, artists, designers … • Engineers or manufacturers from all industries …
  60. 60. Mindset: love it!• Enjoy the thrill of getting a SIGGRAPH paper• Even a bit of an addition• Show joy, not bitterness• Be optimistic
  61. 61. Mindset: patience• Which is harder? • Beautiful and polished images/results/videos • Brilliant presentation • A cool and new idea • Comprehensive evaluation• Luck can lead you to an idea, but not the others!• For those, you need A LOT of patience
  62. 62. Mindset: have fun• “Fun with shapes”• Have your family enjoy it• Buy more time on your submission
  63. 63. TIPS III• Make SIGGRAPH papers your love• Find the right people, mindset, and problem• Keep exploring the more unknown • Shape understanding • Creative modeling and design• When writing, try really hard to make your point
  64. 64. But really …• There is no single recipe To think there is a single type of problem that will make SIGGRAPH is like thinking there is one type of people who is going to be the love of your life!  my quote imitated from Edgar Dijkstra• It is about you … Do only what only you can do!  Edgar Dijkstra
  65. 65. Do what you love• All other factors are ephemeral • Trend, popularity, hotness, …• More likely to be productive and successful • You will spend a lot of time on your stuff• Less likely regret in the worst case (e.g. reject) • At least you have fun• Start your own stuff • It is like investing; followers are already late• Life is too short
  66. 66. Do the best• Graphics → SIGGRAPH • Vision → CVPR, ICCV • …• Hard work anyway; so go for the jugular• Happier if succeed, less sad if fail• Life is too short
  67. 67. Randomness• Humans prefer certainty• Life is random end end start start fiction reality
  68. 68. Randomness in accept/reject clear accept accept borderline deterministic quality bar stochastic quality zone reject clear reject fiction reality
  69. 69. Monte Carlo Sampling• Life long intrinsic acceptance rate r = x% • r seems 0 if the first paper got rejected • r seems 1 if the first paper got accepted • Need more samples! • (be patient, and try more.)
  70. 70. Objectivity• Humans are biased Score by others Score by you • Optimistic → self • Pessimistic → others• Get feedbacks • Early & frequent• Self criticism Paper by you Paper by others
  71. 71. Don’t get mad• (a few) nasty reviewers might exist• Useless to get upset• Get even! • Assume reviewers are going to kill your paper
  72. 72. Discipline• Humans like to procrastinate• Start early• Manage projects by • Paper draft • Schedule
  73. 73. Practice• Humans are lazy• Research is a craft; learn through practice code talk read write experi- create ment
  74. 74. Practice what?• A chain is only as strong as its weakest link • Practice the weakest link Cause of rejection
  75. 75. Be happy• Long term productivity depends on happiness• Live healthy and happy • Sleep, exercise, eat, social life, …• Creativity depends on happiness • I got all my ideas outside office• Be nice and positive to others  • Especially in conferences & reviews
  76. 76. WELCOME TO SIGGRAPH 2013 You won’t be in time if you haven’t written your paper forDanny Cohen-Or SIGGRAPH/Asia 2012… Baining Guo Liyi Wei Olga Sorkine Kun Zhou Hao Zhang