Can anyone tell me if it is possible to use mxnet with matlab on windows? Using the windows packaged releases from https://github.com/dmlc/mxnet/releases, there is no "matlab" directory, which I can see is present in the newer versions, which do not have windows packages.
is there a way to get the best of both worlds?
Yes, it works, even thought I doubt it has been extensively tested. Their Matlab wrapping assumes the library lies in the lib directory, which is inconsistent with their python wrapping that assumes it lies in the build directory. So you will either have to copy libraries or change paths in callmxnet.m. Also, the demo needs data that is currently fetched by a bash script -- not very Windows friendly but it runs flawlessly with MSYS2.
IMHO the Matlab demo is currently more a proof-of-concept than a real Matlab wrapper. It is nowhere near the python bindings.
The matlab directory is directly on the repo not as a release:
https://github.com/dmlc/mxnet/tree/master/matlab
Maybe you can compile it for windows using cygwin or msys2, On the documentation it say that it was tested for matlab version:
Matlab binding has been tested with the following version:
R2016b (9.1.0.441655) 64-bit (glnxa64)
Which is a linux version; I think that it should be posible to reproduce the instructions in the readme to make it work woth cygwin or msys2 with windows
Related
I have download that code from github.
Whenever that code tries to check mexopencv (a file inside lib folder). It shows an error on Matlab 2017a that is:
[error 3] please make sure gcc >= 4.4 g++ >= 4.4 is installed ...
I am having a gcc version of 8.8.0
Following is the description of system
Window 10
Visual Studio 2012
Matlab 2017a
opencv 2.4.10
mexopencv (install from your folder LIB)
You might lack of standard libraries (.dll in Windows but .lib in Linux). The code was developed under Linux (help is also only provided for Linux systems) and once tested under Windows 7.
The method was tested under OS linux lubuntu 14.04, CPU intel i5 and
16GB of RAM. This guide was written under a linux-based distribution
scope, more precisely debian. Unfortunately we lack experience to
address bugs or any problem that could emerge in non-linux OS, we will
provide only linux OS help regarding execution problems, we apologise.
we tested this code on Windows7. We successfully installed opencv
2.4.10 following this tutorial: [http://www.learnopencv.com/install-opencv3-on-windows/]
This suggests that the mex-function is also linux-based. The error should vanish if you compile the source code yourself on your system, e.g. with MinGW (also GCC), see here.
EDITED:
the README.markdown file (under LIB/mexopencv-master.zip) provides a little guide through the installation process of both, openCV and the mexopencv function for unix (including ubuntu 14.04) and Windows systems. The latter states:
Browse to mexopencv root folder, and type the following in the MATLAB
command window:
>> mexopencv.make
By default, mexopencv assumes the OpenCV library is installed in
C:\opencv. If this is not the case, specify the path as an argument:
>> mexopencv.make('opencv_path', 'c:\your\path\to\opencv')
Note that if you build OpenCV from source, this path specification
might not work out of the box. You need to replace dll files in the
OpenCV package with newly built binaries. Or, you need to modify
+mexopencv/make.m to correctly link your MEX-files with the library.
To remove existing mexopencv binaries, use the following command.
>> mexopencv.make('clean', true)
It requires a standard C++ compiler supported by Matlab (see here), any Visual Studio installation should be OK, as an open-source option MinGW is recommended. Just remember to pick the right system (usually 64-bit).
I'm using Matlab R2016a. I'm following this video to compile my Matlab guide program as standalone application. However I can only see options to compile it for a Linux platform. Is it also possible to compile it for Windows using the Linux version of Matlab?
Sadly, at present, the answer to your question is no. In Matlab you cannot achieve cross compilation of executables. The Matlab Compiler is strictly tied to the machine environment.
The basic rule is that in order to build an executable that can run under operating system X, it must be compiler under operating system X.
Therefore, you need to install the Matlab Compiler on a Windows machine to create an executable that can run under Windows. On the top of that, Windows represents a special case in which you have to consider the difference between 32 and 64 bit versions. For more infos, read these threads:
https://it.mathworks.com/matlabcentral/answers/4649-deploy-for-windows-32-bit-from-a-64-bit-computer
https://it.mathworks.com/matlabcentral/answers/217297-32-bit-exe-compilation-using-matlab-64-installed
https://it.mathworks.com/matlabcentral/answers/60904-compile-executable-for-64-bit-system
I have looked at many resources. I just want a simple way to use the newest OpenCV with Matlab 2017a. MinGW doesn't work and a compiler but I can't seem to get Visual C++ working. I can use OpenCV in Visual Studio but what compiler does Matlab want from me and how do I obtain it?
There is a way to use opencv from matlab without compiling any mex files, and without using the computer vision toolbox.
You can call opencv through matlab via python commands.
1) Download and install the CPython distribution, not Anaconda
2) from the command prompt, install the opencv and other packages you want using the pip command > python -m pip install opencv-python
3) Follow the procedure in the matlab help for ensuring matlab knows where to look for the python installation
4) use the syntax py.cv2.function to call the opencv function
If you keep the data types as python data types in matlab while using the py.cv2 functions there isn't too much of a speed hit.
As #Miki mentions in the their comment, have you checked out this page on the MathWorks site? It has many resources, the main one is the Computer Vision System Toolbox OpenCV Interface. It comes with some examples which would be a good starting point. And if those aren't close to what you want, try another one of the community examples on the File Exchange and trying to make small changes to it.
For each version of MATLAB you can find the supported compilers in the documentation. You can also use -v flag as shown blow to see all the paths where MATLAB looks for a C++ compiler.
mex -setup -v
I have problems running a program compiled with MCC that uses parfor. The non-compiled .m version works (no bug). When I run the compiled version, I get
"distcomp.remoteparfor" is undefined. I believe I exactly have the error described in
the following link.
http://www.mathworks.com/support/solutions/en/data/1-PAHWE/index.html?product=CO&solution=1-PAHWE
However, since MCR is not installed on my machine (got Matlab and MCC),
I am wondering what the pathes $APPNAME_mcr/java/jar/toolbox and
$MCR/MATLAB Component Runtime/v70/java/jar/toolbox correspond to. I found
the distcomp.jar file in C:\Program Files\MATLAB\R2011b\java\jar\toolbox, but I don't know where to put it since see any path corresponding to $MCR/MATLAB Component Runtime/v70/java/jar/toolbox since MCR is not installed.
Thanks a lot!
Fred
No need to install the MCR if you are simply trying to run the application on the same machine that has MATLAB. The question to ask is, HOW are you launching the compiled application? From a DOS command window? From within MATLAB using the SYSTEM command?
If you are launching the application from a DOS / UNIX command line, then the important thing to keep in mind is that the MATLAB binaries need to be on your system path i.e. just make sure:
$MATLABROOT\bin\$ARCH
is on your system path. Where $MATLABROOT is the matlab installation folder, and $ARCH is your system architecture. For example:
c:\work\matlab\bin\win64
on my machine since my installation folder which contains MATLAB is:
c:\work
and i am on a win64 machine.
The problem is solved. The solution is to make the program
a function instead of a script :
http://www.mathworks.com/matlabcentral/answers/22825-parfor-errors-when-file-is-compiled
I seem to have wiped out my path environment variable a while back. I've been slowly restoring things. I have both MATLAB and the MATLAB compiler installed. DLL's I've created from the MATLAB compiler won't run because they can't find the MATLAB compiler runtime dlls. While I did find the particular files that are my immediate problem. I'm wondering if there is a MATLAB *.bat file or command I can type that will restore my path variable to what it was after MATLAB and the Compiler were installed. I'm hoping to forestall future problems.
In case it's version specific I'm running MATLAB R2010b, I'm running a 32 bit version on a 64 bit machine.
[Edit]
I thought I would add that the path I need for running the compiled version was:
C:\Program Files (x86)\MATLAB\MATLAB Compiler Runtime\v714\runtime\win32
With luck, that one along with the two suggested in the answer will get me back to the original state.
restoredefaultpath might recover your MATLAB installation. Consider the use of startup.m, in order to easily undo changes to your environment.
You want to have these two directories on the PATH (I think the order is important):
C:\Program Files\MATLAB\R2010b\runtime\win32
C:\Program Files\MATLAB\R2010b\bin
Obviously you need to adjust the path to match your setup and architecture (those are on a WinXP 32-bit)
If you have a current software maintenance contract for your MATLAB, it may well be that the easy thing to do is just upgrade to the next version of MATLAB since R2011a is out now (unless there is some compelling reason why you must develop on R2010b. Running the software installer should recreate the default environment for MATLAB. I am, of course, assuming that you have Administrator access on your PC and have permission to install software.