[From official texts]
Monday 12 December | 09:00-12:45 | Room S222
With the advancement in 3D display technologies, we anticipate that the next move in the display industries will gear toward autostereoscopy for multiple users. Such a new paradigm will ask for a novel approach to acquiring, processing and synthesizing a real 3D scene at arbitrary viewing directions. As one of the candidates for future 3D imaging technology, Time-of-Flight (ToF) depth camera receives great attention from various researchers and has been adopted in several topics. In this tutorial, we introduce the recent advancement of 3D depth sensing technologies and provide its potential use in 3D imaging for future 3D display. Motivated by recent progress in depth image processing and inverse rendering algorithms, we organize this timely and unique tutorial to introduce basic principles, in-depth discussion on the cutting edge of technical issues, potential ideas, challenges. This tutorial will provide well organized presentations on: the principle of ToF Depth sensor and its sensing architecture issues; the state of the art depth processing algorithms; 3D reconstruction from color and depth images and the lighting and reflectance extraction from color and depth images.
[From official texts]
The course will challenge the simple question of "how to write a SIGGRAPH paper". The course will cover a larger scope than just how to "write" the paper, but it will consider many other aspects: how to pick a good problem, how to position it, how to produce convincing results, how to evaluate it, how to present the results, etc. In the course, the presenters will share their experience with the audience hoping to provide them with useful tips and knowhow that will help them in realizing the potential of their research. The course will consist of two main parts: the first three senior presenters will give their personal view, and in the second we will have a larger panel that includes a few younger successful researchers.
16:10-16:30 Lecture 3: How to write a SIGGRAPH paper, Dani Lischinski
SIGGRAPH Review form
Are the fererences adequate?
List any additional references that are needed.
- Did you cite/discuss any work that the reviewers might feel to be relevant?
- Think who might be chosen
Could the work be reproduced from the information in the paper? Are all important algorithmic or system details discussed adequately? are the limitations and drawbacks of the work clear?
- completeness: mention important implementation details, constants, parameter values.
- Make sure your demonstrate/discuss any drawbacks or limitations! Almost nothing works 100 percent.
Who cares about the introuction?
- The importance of the introduction cannot be overestimated!
- Uneducated guess: in over 90 percent of the cases, the reviewer will have made up his mind, while reading the introduction.
- What is this paper about?
- What problems does it address?
- Why should the reader/reviewer caer?
-- Convince me the problem is important
-- Convince me it has not yet been solved well enough
-- Convince me have a novel solution
-- Make me want to read...
Possible Introduction Structure (I)
- Starting sentence/paragraph: introcude the broader context for this work. exlainin its importantce ( as general motiation).
- Next paragraph or two: narrow the context down. State...
Hasn't this problem been adequately solved by previous research? Breafly state why the exiting approaches still leave room for this work. This sets the cround for stating the contribution of this paper.
1: Opeining paragraph" (context + motivation)
2: Narrowing down
3: Summarize the approach presented in this paper
4: Any limitations to declare?
- Make good use of your figures!
-- Demonstrate the problem you are solving
-- Show the shortcomings of existing methods
-- Visual aids to help explain your approach
-- Demonstrate the quality of your results
- Ideally, the reviewer should understand the poin of your paper just from...
Tps -Keeping Promises
- Maintain a balance between your claims (promises) in the introduction, previous work, etc...
Tips - Previous Work
- Be through
- Be kind/fair
- Support your claims about shortcomings
- Don't just write a "laundry list"
Tips - Method
- Provide a high-level overview: a logical "road map" for the exposition of your method
- Expose in a top-down...
Tips - Results
- Don't assume the reader will understand on his own what makes your result better than the existing state-of-the-art
Tips - conclusions
-Summarize what you have (but also what you have not) archived.
16:30-18:00 Panel which include a number of successful researchers, TBD
Olga Sorkine (OIGL/ETH)
Who to write a serious SIGGRAPH paper and Get Away with it.
- Tries to solve a problem you deeply care about.
- The problem might not be new (but your solution should be!)
- Contains highly nontrivial algorithms and possibly...
- It is (sometimes) more difficult to improve upon exitsting work than to work on a new problem.
(color Harmonization, SIGGRAPH 2006)
-Must avoid the "delta" impression
- Your solution requires deep math understanding, and your formulas are tough.
- Elucidate as much as you can
- Chrystal clear explanations
- Intuition is always helpful
- Dont't try to hand-wave and trivialize things away. But try to find the simplest ... (photo)
Practical Tips for Math Haircut
- Clear and neat notation
- Always define all symbols
- Give equations numbers
- Assume nithing - explain everything
How to get away with it
- Make sure you speak the language of the community
- Do your homework - learn...(photo)
Terminology is Important
-Example: "Linear Rotation - Invariant Coordianates" paper (SIGGRAPH 05)
-Original title: "Bonnet Mesh Manipulation
-Renvented discrete fundeamental forms
-...badly made it in and had to do a lot of rewriting
-Make your readers' (reviewrs') life as easy as you can.
-Main thing : aim at innovative, impactful research work
-SIGGRAPH, too, can make mistakes. So:
Don't discare (??あきらめないで）
Write three first-author SIGGRAPH papers in one year - tips for graduate students
"Kun Zhou" Zhejiang University
My experience #1
Right ideas: novel, not incremental
Read as many papers as possible
- Attend SIGGRAPH
- Geometry, texture ,rendering papers
- SIGGRAPH demos
Exmaple: shadow fields
- Problem; extend PRT to dynamic scenes
- Inspired by a collision detection demo
"BD-Tree", SIGGRAPH 2004
DougL. James and Dinesh K. Pai
My experience #2
Results/Demos: technically awesome
- Sometiong the state-of-the-arts cannot do
- Comparisons are often unavoidable
My experience #3
Writing: get help and learnd from experienced people
- Shadow fields: Stive Lin, Baning Guo, Harry Shun
- VGL demormation: John Snyder
- Texture Montage; Matthieu Desbrun
- Single idea 4 pages
My experience #4
Schedule your project as early as possible
-Example : Shadow fields
-- Initial idea, May 3004
-- Proof-of-concept implementation, Aug. 2004
-- Efficient implementation, Nov. 2004
My experience #5
The best of luck
- SF 4.0 4.2 3.7
My experience #6
Take are your ...(photo)
Start trying when your are 22-25 years old
-Idealistic: beatiful. even sexy, pretty name
Do not sue me...
Be a little bit careful if you are married
Enjoy the thrill of getting a SIGGRPH paper
What is graphics?
Creation, display, storage, and animation of visual content
Cration of novel
in his talk
"Towords the first SIGGRAPH paper"
Hao (Richard) Zhang