drakeguan.org

2011 Python 回憶錄

Drake — Wed, 11 Jan 2012 04:54:10 +0000

2011 走了，民國 100 年在眾人互相廝殺的 boardgame, 美味湯頭火鍋, 以及散步看 101 煙火去中度過；2012 世界末日年也來了，我覺得 2012 年比其它年幸運，大家都記得它有個世界末日的外號，其它年份未必有這個福氣被大多數人記著。

論語《為政》篇裏頭提到“溫故而知新，可以為師矣。”參考 Tarek 的計畫，趁勢來個 2011 Python 回憶錄。

1. What's the coolest Python application, framework or library you have discovered in 2011?

virtualenv, pip, Fabric, logbook, and mongoengine.

2. What new programming technique did you learn in 2011?

map reduce, git, optimization by matlab, and TDD.

3. What's the name of the open source project you contributed the most in 2011? What did you do?

vimrc, partio.

4. What was the Python blog or website you read the most in 2011?

I think it's PyPI.

5. What are the three top things you want to learn in 2012?

git, pyqt (restart), ipython(rebrush), matplotlib and mobile programming.

6. What are the top software, app or lib you wish someone would write in 2012?

high-quality combinatorial renderer on cloud?

Refer to:

Daily:

python

Pixar's RenderMan in SIGGRAPH ASIA 2011

Drake — Sat, 03 Dec 2011 06:39:09 +0000

After spotting this exhibitor DM, it reminds me of one un-answered question:

Do we really realize the contribution and advantages of Pixar's RenderMan? If we don't, how can we evaluate other renderers (ex, Arnold Renderer) in that soon without any hesitate?

Daily:

siggraph

pixar

renderman

SIGGRAPH ASIA 2011: Technical Papers Trailer

Drake — Sun, 30 Oct 2011 16:53:12 +0000

The followings are the research titles in the trailer video with html copied from Ke-Sen's work: SIGGRAPH Asia 2011 papers on the web.

Rendering Synthetic Objects into Legacy Photographs: Kevin Karsch, Varsha Hedau, David Forsyth, Derek Hoiem (University of Illinois at Urbana-Champaign)
Modeling and Generating Moving Trees from Video: Chuan Li (University of Bath), Oliver Deussen (University of Konstanz), Yizhe Song, Phil Willis, Peter Hall (University of Bath)
Video Face Replacement: Kevin Dale, Kalyan Sunkavalli (Harvard University), Micah K. Johnson, Daniel Vlasic, Wojciech Matusik (Massachusetts Institute of Technology), Hanspeter Pfister (Harvard University)
Candid Portrait Selection From Video: Juliet Bernstein (University of Washington), Aseem Agarwala (Adobe Systems, Inc.), Brian Curless (University of Washington)
C1x6: A Stereoscopic Six-User Display for Co-located Collaboration in Shared Virtual Environments: Alexander Kulik, Andre Kunert, Stephan Beck, Roman Reichel (Bauhaus-Universitat Weimar), Roland Blach Fraunhofer IAO (Digital Image), Bernd Froehlich (Bauhaus-Universitat Weimar)
A Hybrid Iterative Solver for Robustly Capturing Coulomb Friction in Hair Dynamics: Gilles Daviet, Florence Bertails-Descoubes, Laurence Boissieux (INRIA, Grenoble, France)
Interference Aware Geometric Modeling: David Harmon (New York University), Daniele Panozzo (University of Genova), Olga Sorkine (ETH Zurich), Denis Zorin (New York University)
Controlling Physics-Based Characters Using Soft Contacts: Sumit Jain, C. Karen Liu (Georgia Institute of Technology)
Animated Construction of Line Drawings: Hongbo Fu (City University of Hong Kong), Shizhe Zhou, Ligang Liu (Zhejiang University), Niloy Mitra (King Abdullah University of Science and Technology)
Mixed-Order Compositing for 3D Paintings: Ilya Baran, Johannes Schmid, Thomas Siegrist, (Disney Research), Markus Gross (ETH Zurich), Robert W. Sumner (Disney Research)
Insitu: Sketching Architectural Designs in Context: Patrick Paczkowski, Min H. Kim, Julie Dorsey, Holly Rushmeier (Yale University), Yann Morvan, Carol O'Sullivan (Trinity College Dublin)
Displacement Interpolation Using Lagrangian Mass Transport (TR): Nicolas Bonneel, Michiel van de Panne (The University of British Columbia), Sylvain Paris (Adobe Systems Inc.), Wolfgang Heidrich (The University of British Columbia)
Sketch-based Dynamic Illustration of Fluid Systems: Bo Zhu (ERATO, JST), Michiaki Iwata, Ryo Haraguchi (National Cerebral and Cardiovascular Center Research Institute), Takashi Ashihara (Shiga University of Medical Science), Nobuyuki Umetani (The University of Tokyo), Takeo Igarashi Kazuo Nakazawa (National Cerebral and Cardiovascular Center Research Institute)
Modular Radiance Transfer: Bradford J. Loos (Disney Interactive Studios), Lakulish Antani (University of North Carolina) , Kenny Mitchell, Derek Nowrouzezahrai, Wojciech Jarosz, (Disney Research Zurich), Peter-Pike Sloan (Disney Interactive Studios)
AppGen: Interactive Material Modeling from a Single Image: Yue Dong (Tsinghua University), Xin Tong (Microsoft Research Asia), Fabio Pellacini (Dartmouth College / apienza University of Rome), Baining Guo (Microsoft Research Asia)
Genetic Programming for Shader Simplification: Pitchaya Sitthi-amorn, Nicholas Modly, Westley Weimer, Jason Lawrence (University of Virginia)
A Shading Reuse Method for Efficient Micropolygon Ray Tracing: Qiming Hou, Kun Zhou (Zhejiang University)
Progressive Photon Beams: Wojciech Jarosz, Derek Nowrouzezahrai, Robert Thomas (Disney Research Zurich), Peter-Pike Sloan (Disney Interactive Studios), Matthias Zwicker (University of Bern)

Daily:

siggraph

research

paper

OpenHealth 2011 工作坊

Drake — Sun, 30 Oct 2011 16:01:35 +0000

OpenHealth的工作坊特邀曾任教於 MIT Media Lab（Context Aware Computing Group），現任教於卡內基美隆大學的教授Ted Selker，以探索健康生活為主題，歡迎對互動設計與智慧運算有興趣學子們，與大師共同激盪出有趣、有創意的健康生活新方案。

本工作坊以「健康生活」為概念下，延伸出四個主題：「居家」、「飲食」、「城市」與「照護」。每個主題會由不同助教所引導，利用各種有趣的腦力激盪方法，讓不同領域的學員們共同激盪出有創意的概念與互動原形，並結合自己擅長的技術與專長，在密集的工作坊中發揮最大的潛力。

More info on the official webpage.

Daily:

event

hci

design

National Taiwan University in SIGGRAPH Asia 2011

Drake — Wed, 26 Oct 2011 15:43:32 +0000

Congratulations on those guys (including me) :)~

Face Recognition and Clustering for Home Photos

In this work, we focus on clustering faces in home photos with face recognition technologies.

Che-Hua Yeh, National Taiwan University
Pei-Ruu Shih, National Taiwan University
Kuan-Ting Liu, National Taiwan University
Huang-Ming Chang, National Taiwan University
Ming Ouhyoung, National Taiwan University

Stereoscopic 3D Experience Optimization Using Cropping and Warping

The purpose of this paper is to optimize the stereoscopic 3D experience. We propose ﬁve principles to reduce visual fatigue and enhance depth perception by cropping and warping.

Hong-Shang Lin, Taiwan University
Chu-Tien Lee, Taiwan University
Shuen-Huei Guan, Digimax Inc.
Ming Ouhyoung, Taiwan University

src: Technical Sketches | SIGGRAPH ASIA 2011 HONG KONG.

我看 Upright Orientation of Man-Made Objects

Drake — Tue, 11 Oct 2011 14:58:44 +0000

Siggraph 2008 的一篇論文：Upright Orientation of Man-Made Objects。這篇論文的目的是要把 3D 模型給擺正。只要你隨意給一個 3D 模型，使用這篇論文裏的方法的話，就能把一個模型擺正到可以平放在桌面上來。（註：其實是人造的物件為主，如果你是給一隻兩隻站立的狗的話，不曉得會跑出什麼結果來~）

影片觀後感：

PCA 很可憐，只要是涉及 domain knowledge 的東西，都會故意拿出 PCA 當作對照組，然後它就會被數落一番...
顯微鏡好用，一路狂用來打 PCA...
小朋友的那段很優，想到用小朋友來示範『把東西擺正』的這件事，實在很有意思。
開場的機器人站起來那段，整個就很驚人，我猜他們看到這結果時，應該非常興奮，就整個覺得 paper 會上了吧。

事後觀點：

放寬 upright 的定義，就可以用在很多地方?
挑對 domain attributes 的話，supervised learning by Machine Learning 挺厲害。ML -> 不是發現新事物，而是定義出看待事物的規則來。
整篇論文的實作不易，都是描述性的文字敘述
Google Scholar 結果。

Daily:

Pixar 用動畫來教你 RenderMan 是怎麼運作的

Drake — Mon, 03 Oct 2011 01:53:02 +0000

今年的 Stupid RenderMan/RAT Tricks 2011，Pixar 的 Patrick Coleman 給了一個簡短的 trick: Illustrating RenderMan for the Masses。整個 talk 的目的圍繞在一個主題上：

跟你的奶奶解釋 RenderMan 是怎麼協助電影動畫的。

上頭這個副標不錯吧，也很難吧。事實上，你把上頭的 RenderMan 換成任何一個你熟悉的技術或東西，都一樣困難。試試換成 mp3、電子書、網際網路、Facebook、iPad、App…以我自己來看的話，我想 mp3 是最容易解釋的一點 XD

上頭的影片短短的，但可以稍為搭配一下放出來的投影片，裏頭至少有幾個關鍵字，包括像是 bucket, grid, shading，好像就這樣。趕緊拿去試試看，怎麼向你的奶奶解釋你平日的工作內容是什麼吧~

Daily:

pixar

renderman

video

Renderings 101 for videogames

Drake — Sun, 02 Oct 2011 03:09:43 +0000

I found this slides accidentally and it reads fluently and describes renderings clearly. If I had to give a course of semester about rendering, I might borrow this slides and address it in the 1st or 2nd hour of the course. It should be a good test to make participants to decide to enroll this course or not asap.

(videogame) Rendering 101

Summary of SIGGRAPH 2011 in Prezi

Drake — Wed, 31 Aug 2011 05:47:26 +0000

My personal viewpoints of SIGGRAPH 2011 after lots of discussion with friends, colleagues, and someone. It is very subjective and the presentation on Prezi is just a testing. The presentation is not completed yet but I think it's enough to leave it in this way. Check the source presentation: Summary of SIGGRAPH 2011 on Prezi

Daily:

siggraph

prezi

presentation

Summary of SIGGRAPH 2011

Drake — Fri, 26 Aug 2011 15:15:59 +0000

More and more research papers focusing on controllable or guidable topics are proposed, like ShadowDraw, NPR Noise, Example-Based Elastic Materials, Photo-Inspired Modeling, Assembly-Based 3D Modeling.
Alembic 1.0 is proposed to supersede long-running OBJ and work successfully in this industry, just like OpenEXR for images.
Open source might not be the right way, but joint development is the way!
Deep image compositing(1, 2, 3) is hot. Weta developed and used it for Avatar on Pixar's Renderman and The Foundry's Nuke is going to support that. Furthermore, deep image format is included in the standardization of OpenEXR 2.0.
Katana is going to be release in this year, which is a 3D application specifically designed for the needs of look development and lighting in an asset based pipeline.
The course: Stereoscopy from XY to Z is awesome and really informative. ( slides by NVIDIA, brief on real-time rendering)
Global illumination by ray tracing is the trend but volumetric rendering is the next!
Shrink Wrapping, low-res and distant cars in Cars 2 are modelled by a simple cube with displacement shaders. Not a new idea but it still looks cool.
When we talk about 3-D, the stereoscopic vision, people are still mentioning the min/max disparity/parallax issues, like some safe zone, the bearable disparity angles, and the min/max stuff. Even Disney puts 'comfortable' as the top of stereoscopic principles. But, Disney also starts to mention emotional stereo and emotional distance.
Current volumetric renderers still base on ray tracer or REYES.
Blue Sky thinks crowd simulation takes time, slower loading and easy to crash your memory-limited machines, it is better to just bake variety of animations and use that to synthesize plausible crowd animation. Well, that is all because Blue Sky is using proprietary ray-tracing renderer! But in some situation, it makes sense.
Sparse Voxel Octree (SVO)(google code: Efficient Sparse Voxel Octrees) is the key to LoD.
CG is about cheating and we all should know that in advance.

Daily:

siggraph

Links for 2010-08-25 [del.icio.us]

Thu, 26 Aug 2010 00:00:00 PDT

Mars 5D 's Blog - ! Ù

Links for 2009-11-14 [del.icio.us]

Sun, 15 Nov 2009 00:00:00 PST