GitXplorerGitXplorer
m

matlab-opencv-interop

public
2 stars
2 forks
0 issues

Commits

List of commits on branch master.
Unverified
b650ff49e6e21521e49c7bfa52bf6157458a79ae

rf_examine_leaves done but doesn't run, needs debugging.

mmalcolmreynolds committed 13 years ago
Unverified
b3e9434d94288ac8d23920f3ac8cc1a5cd9a69e0

Started rf_examine_leaves.cpp

mmalcolmreynolds committed 13 years ago
Unverified
c60c990d7ef21d2c63a49d4006b11467443282e9

First smart training system, does cross validation for param search.

mmalcolmreynolds committed 13 years ago
Unverified
1e28c8a6a17a281e0d370b85e3ce50cc347ed731

Added MEX_ERR_PRINTF macro, removed tonnes of boilerplate

mmalcolmreynolds committed 13 years ago
Unverified
54ea12d7ef056ec9f8c6f48dcc94102f83af6afa

Merge remote-tracking branch 'prism/master'

mmalcolmreynolds committed 13 years ago
Unverified
88fdb5ec7ce94651d9553303895021c451f04580

Various windows fixes, including one heck of a crash.

mmalcolmreynolds committed 13 years ago

README

The README file for this repository.

OpenCV bindings for Matlab - Readme

malcolm.reynolds@gmail.com

This is a simple set of bindings for a very small subset of OpenCV functionality so that it can be called from Matlab. I wrote these bindings for just the bits of OpenCV I needed for a specific project, which was mainly the Random Forest implementation. If you need bindings for lots of other parts of the system, it's probably time to switch to Python like I'm doing.

Anyway, the included Makefile works on OS X (64 bit) and with suitable modifications to the configuration to the bits at the top I'm sure it can be made to work for any platform that matlab runs on.

Requirements:

  • Matlab: Works on 2010rb and 2009aSV. In principle any version that includes MEX support should be fine.
  • OpenCV: Tested on OpenCV 2.x. This library seems to get reorganised fairly frequently so some headers / includes might need changing at some point...
  • C++ compiler: tested with GCC, but whatever Matlab wraps with the 'mex' command should be fine.

Matlab needs to be installed such that you can type "mex" to run it in a terminal. The actual file should be in your matlab installation in the bin/ directory.

OpenCV's libraries and include files must be given to the makefile - I use a shell script to print out the locations of this (in -I/some/include/directory -L/some/library/directory -lfirstlibrary -lsecondlibrary format), which is the opencv-comp in the makefile. You can just manually enter them here too if you like, but if you ever do any other opencv compilation on the command line the shell script is handy.

One you're compiled you should have a bunch of .mexw32 / .mewx64 / .mexmaci64 / whatever files sitting in the same directory as the source files. Point Matlab's PATH to them and you're good to go.

Disclaimer

This was written for a single project, and it worked for me. No guarantees - I saw some particularly gnarly crashing bugs on Windows XP 32 bit when I ported it over, but I simply do not have the inclination to find out what's going on there. I was planning to expand this to cover more opencv functionality, but I am switching my workflow to Python/Scipy/Numpy for the future. However, this is here just in case anyone wants simple, high performance, reliable random forests in matlab.

All complaints to malcolm.reynolds@gmail.com - if you use this for something useful, I would love to hear about it.