caffe_.mexw64: Invalid MEX-file about faster_rcnn HOT 21 CLOSED

The caffe_.mexw64 was downloaded from dropbox link.

from faster_rcnn.

Please check whether system environments are compatible with the downloaded mex (including cuda 6.5, vs2013, windows 7/8).

from faster_rcnn.

Almost the same thing for me too (but for Ubuntu OS):

Mex file entry point is missing. Please check the (case-sensitive)
spelling of mexFunction (for C MEX-files), or the (case-insensitive)
spelling of MEXFUNCTION (for FORTRAN MEX-files).
Invalid MEX-file '/home/agruzdev/git/faster_rcnn/external/caffe/matlab/+caffe/private/caffe_.mexa64':

Error in caffe.get_net (line 27)
hNet = caffe_('get_net', model_file, phase_name);

Error in caffe.Net (line 33)
self = caffe.get_net(varargin{:});

Error in script_faster_rcnn_demo (line 33)
rpn_net = caffe.Net(proposal_detection_model.proposal_net_def, 'test');

The caffe_.mexa64 was build from the latest version of caffe for faster r-cnn.

from faster_rcnn.

@AlexGruzdev For Ubuntu, the compliing phase is the same as official Caffe.

from faster_rcnn.

@ShaoqingRen I didn't download caffe_.mexa64 file, I did compilation according official Caffe instructions.

from faster_rcnn.

@ShaoqingRen I have same problem as @AlexGruzdev...

configured using cmake, after trying to load caffe_:
using debian jessie, matlab R2014a:

caffe_
Mex file entry point is missing. Please check the (case-sensitive)
spelling of mexFunction (for C MEX-files), or the (case-insensitive)
spelling of MEXFUNCTION (for FORTRAN MEX-files).
Invalid MEX-file '/home/smichalowski/faster_rcnn/external/caffe/matlab/+caffe/private/caffe_.mexa64':

from faster_rcnn.

I haven't been able to reproduce this problem on Linux. I'm using matlab R2014a, gcc 4.8.4, ubuntu 14.04.3.

@AlexGruzdev, @smichalowski When you cloned the faster_rcnn repo, did you use git clone --recursive? If not, please try that to make sure that you're on the correct commit for the caffe submodule (d1d96fe).

from faster_rcnn.

hi @rbgirshick

Yes, I cloned faster_rcnn with --recursive option.
Today Ive tried ubuntu 15.04 + gcc 4.7 + matlab R2014a - same error.
caffe was built using cmake, I'll try again with standard make and Makefile.config

from faster_rcnn.

The issue could be with cmake. I used make and this Makefile.config.

from faster_rcnn.

The issue is most likely that the cmake build is not defining BUILD_MEX_INTERFACE. Makefile was modified to add -DBUILD_MEX_INTERFACE to the CXXFLAGS.

@ShaoqingRen, was there a reason for adding the BUILD_MEX_INTERFACE guard? Perhaps it should just be removed from caffe_.cpp.

from faster_rcnn.

@rbgirshick BUILD_MEX_INTERFACE is used in VS solution, I'll make BUILD_MEX_INTERFACE only affect in Windows.

from faster_rcnn.

@AlexGruzdev @smichalowski The BUILD_MEX_INTERFACE only functions on Windows now. Please have a try with the new version.

from faster_rcnn.

yes, compiled without errors. thank you

from faster_rcnn.

With your last version of caffe I got the following:

Building with 'g++'.
/home/agruzdev/git/faster_rcnn/external/caffe/matlab/+caffe/private/caffe_.cpp:19:28: fatal error: gpu/mxGPUArray.h: No such file or directory
compilation terminated.

make[2]: *** [matlab/+caffe/private/caffe_.mexa64] Error 255
make[1]: *** [matlab/CMakeFiles/matlab.dir/all] Error 2
make: *** [all] Error 2

from faster_rcnn.

@AlexGruzdev - please use the Makefile build; the cmake build is unsupported in this fork. We updated the Makefile to add the following include path and lib for compiling the matlab wrapper:
MATLAB_INCLUDE := -I$(MATLAB_DIR)/toolbox/distcomp/gpu/extern/include
MATLAB_LIBS := -lmwgpu
If you wish to use cmake, you'll need to incorporate these additions into the cmake build.

from faster_rcnn.

@ShaoqingRen @rbgirshick Thank you, guys!
Compilation without errors.

from faster_rcnn.

Great. Closing this issue.

from faster_rcnn.

thanks for you guys

from faster_rcnn.

there is a cuda7.5 version in the rfcn github, i download it and set tu cpu mode and it works...

from faster_rcnn.

Please guide me.

My detailed problem link is here.

from faster_rcnn.

how to acquire caffe_.mexa64 under LINUX?

from faster_rcnn.