It’s been a while of searching before I found this one: A Quantitative Analysis of Current Practices in Optical Flow Estimation and the Principles Behind Them, a survey paper on today’s stereo algorithms.

The author provided a collection of those algorithms implemented in Matlab. It’s quite exciting me since there’s no dependece problem any more.

More interesting thing is the author, Deqing Sun postdoctor of Havard, appears to be a Hongkonger, and has been in this area for years of research. Hope soon I can get in contact with him.

I am starting fresh on the project of stereo matching on flow. This is a hard work since Bai did not study in this area, and few resource can be found in our lab. I’ve come through some papers, but the key problem in this point is the program, so that I can test the algorithm and put the study in practice.

After failing to search the code for the #1 KITTI paper View-Consistent 3D Scene Flow Estimation over Multiple Frames., It’s been a while I’m caught in trouble. Hopefully on Friday Oct 10, I ran into some dataset MPI, and eventually found a well-organized resourse, provided by LEAR. It’s amazing that the thing comes to you right when you’ve got nothing to turn to, and the dataset itself, is beautiful and gorgerous, making it outstanding and attractive other than any data I’m come to before.

And that’s how things are going: I’m starting from scratch from the code. The matching algorithm they give is an implementation of deep learning, rather similar to what I’ve talked in the meeting, but stands on a higher level of image matching. A good thing is their code is well organized, so I can build things upon them.

However the compiling stage is not that organized as common project standard asks. It’s been a while to set things right on a new machine, and it makes sense to have a record of the set-up stage.

I’ve planned to use the tool under my Mac, but it anoys me that the code uses many headers like

, that is included by
in later c enviroment, but is not valid under Mac System. So I turned to Ubuntu 14.04 operating as a virtual machine on Vmware Fusion 7.

The ‘README.txt’ in the project does not tell the dependence, only mentioned that the code is compiled right under Fedora. However default Fedora may not serve all the dependence as well. To set up the code, there are some libaries to install. Notice what I’ve done may not perform right on some other machine, depending on the system and hardware.

The compiling enviroment is

   sudo apt-get install gcc-4.8

The media processing needs boost

   sudo apt-get install boost-all-dev

The version is 1.5.4

The image file processing needs

, which is in

   sudo apt-get install libjpeg-dev lib-png12-dev

The linear algorithm needs

, which is in

   sudo apt-get install libatlas-base-dev liblapack-dev

By now I can’t fix the compiling problem for this project; Leave it as undone and maybe somebody can help me with it.

The slides for today’s topic. For offline use link.

Reference for you:

Information is beautiful



Wikipedia edit wars: The most controversial topics

Nobels, no degrees

OrgOrgChart: The Evolution of an Organization

Poetry on the Road 06


CIC China Social Media Landscape 2013

The most unisex names in US history

A visual exploration on mapping complex networks


Linkedin inmaps

It’s an honor for me to be here presenting this topic to you: the newly developed and powerful web design tool & idea, the responsive design.

The most content I’ll cover today is based on the slides, which I picked up from the Internet. The download url is also for ppt and keynote.

Since the slides doesn’t contain much technic details, I’ll introduce some useful tools & websites for your reference.

And also a general introduce following for you.

What is it?

The name was first introduced by ETHAN MARCOTTE in his famous blog Responsive Web Design. However the idea comes from even earlier the year of 2000, when the blog A Dao of Web Design was addressed.

The control which designers know in the print medium, and often desire in the web medium, is simply a function of the limitation of the printed page. We should embrace the fact that the web doesn’t have the same constraints, and design for this flexibility. But first, we must accept the ebb and flow of things

What’s the problem nowadays?

  • Different Platform:
    • Windows
    • Mac
    • Linux
    • iOS
    • Android
    • WP 7
    • WebOS
    • Chrome OS
  • Plenty of devices
  • Multiple resolutions
  • Landscape & portrait
  • Smart devices increasing quickly.
  • Cloud Computing, browser is the future

The Responsive Web Design

  • No more sub domains
  • No more iPhone-special page
  • One page for all devices

How to do it

  • Media query
  • Fluid grid: using percetage length
  • Flex box


  • Less working
  • No more device-special sub-domain
  • Adaptive many kinds of devices


  • Bandwidth limitation
  • Small memory & slow CPU
  • Download useless resources
  • Image size


  • Design for mobile first,pc as an extend
  • Use a fluid grid
  • Modular
  • No iframe
  • Fluid img: 100% width or max-width,so does videos


Reference & Recommends


Camera Exposure Simulation: 一个由佳能提供的在线的曝光模拟

The backstards book of photography: 我非常喜欢的一个摄影教程

纽约摄影学院摄影教材 上册 下册

我的一些照片 欢迎吐槽