caffe_.mexw64: Invalid MEX-file Windows

[PhDStudentC]

i followed the instruction of https://github.com/ShaoqingRen/caffe/tree/faster-R-CNN to create my mex-File( win8.1, matlab 2014a , vs2013, cuda 7.5) but when i tried to test the faster-rcnn demo i had this problem

GPU 1: free memory 2059571200 Use GPU 1 Invalid MEX-file 'C:\Users\Public\faster_rcnn-master\external\caffe\matlab\caffe_fasterrcnn\caffe.mexw64': The specified module could not be found.

Error in caffe.setdevice (line 9) caffe('set_device', device_id);

Error in active_caffe_mex (line 26) caffe.set_device(gpu_id-1);

Error in script_faster_rcnn_demo (line 10) active_caffe_mex(opts.gpu_id, opts.caffe_version);

[YingjieYin]

Maybe you should Run fetch_data/fetch_caffe_mex_windows_vs2013_cuda65.m to download a compiled Caffe mex (for Windows only).

[PhDStudentC]

i traied to use it but the same prob :( (i have cuda7.5 so i have to use my own mex file like in the https://github.com/ShaoqingRen/caffe/tree/faster-R-CNN )

[cominger]

Did you find the way solution to use Cuda 7.5? By the way I have same settings and same problems. However change the Cuda version to 6.5 able to run the demo.

[PhDStudentC]

this problem was caused by the gpu (less than titan ) and some missing dll so the demo run fine on cpu and i am trying to test it with such gpu

[lightning0016]

i'm also cuda7.5,is any success in cuda7.5?thankyou

[PhDStudentC]

hi, yes it works in cuda 7.5 :)

[s0302102]

@ShaoqingRen , @YingjieYin @PhDStudentC I got the same problem. I use cuda 7.0 and compile the caffe_.mexw64 successfully. When I use dependence walker to check what are missed, even there is libmx,libmex(from matlab), and cudart65_64.dll missing (include some other more). But I didn't use cuda 6.5 at all and I couldn't find any environment using cuda 6.5. I don't have installed cuda 6.5 in my computer. If such dlls missing, why can I build it successully?

[shiorioxy]

@PhDStudentC hi，can you send me the mex file which is compile on vs2013, cuda 7.5. Thank you! My email is shiorioxy@qq.com

[jian-ke]

Hi everyone，I use Cuda 7.5，Feforce GT730. The caffe_.mexw64 works fine.

[RichardWenz]

Hi. I'm currently trying to run the code according with the instruction but i got the same error and i don't know how to solve this since i am new in GPU programming. I am using Windows 8.1 with Matlab R2015a, VS2013 and CUDA7.5 on GeforceGTX950M. Is there anyway to fix this problem?Please share if there is any. Thank you in advance!

[RichardWenz]

hi @shiorioxy did you get the mex file?i would like to have it if you got it since we encountered the same problem. Thank you.

[mindcont]

hi @PhDStudentC @ckmufeng I use Window 10 、Matlab R2016a, vs2013, detail like this,

环境	版本
Window 10	Version 10.0.10240
Matlab	R2016a
Visual Studio 2013 [Ultimate]	Update 5
NVIDIA CUDA	7.5.18_WIN10
OpenCV	2.4.9
Boost	1.61.0
显卡(GPU Card)	NVIDIA Quadro K2200

Fllow the step Caffe for Faster R-CNN,I have successfully compiled caffe_.mexa64 ,and Copy all files in .\x64\Release_Mex to faster_rcnn-master\external\caffe\matlab\caffe_faster_rcnn,

than I also got the error like this :

GPU 1: free memory 2059571200
Use GPU 1
Invalid MEX-file 'C:\Users\Public\faster_rcnn-master\external\caffe\matlab\caffe_faster_rcnn\caffe_.mexw64': The specified
module could not be found.

Error in caffe.set_device (line 9)
caffe_('set_device', device_id);

Error in active_caffe_mex (line 26)
caffe.set_device(gpu_id-1);

Error in script_faster_rcnn_demo (line 10)
active_caffe_mex(opts.gpu_id, opts.caffe_version);

Can you tell me how you solved it？

very thankful !!!

[mindcont]

Update:

I have tried on Ubuntu 14.04 to run the fatser_rcnn in CPU MODE and it works for me . So I suggest that when you meet this kind of question can try this method first ，all you need to do is

1 . Edit and compile caffe in $FRCN_ROOT/external/caffe/Makefile.config

# cuDNN acceleration switch (uncomment to build with cuDNN).
# USE_CUDNN := 1

# CPU-only switch (uncomment to build without GPU support).
 CPU_ONLY := 1

and complie by

make -j 4
make matcaffe

2 . Switch to CPU mode by edit $FRCN_ROOT/experiments/script_faster_rcnn_demo.m

%% -------------------- CONFIG --------------------
opts.caffe_version          = 'caffe_faster_rcnn';
% opts.gpu_id                 = auto_select_gpu;
% active_caffe_mex(opts.gpu_id, opts.caffe_version);

opts.per_nms_topN           = 6000;
opts.nms_overlap_thres      = 0.7;
opts.after_nms_topN         = 300;
opts.use_gpu                = false;
opts.test_scales            = 600;

%% -------------------- INIT_MODEL --------------------
%model_dir                   = fullfile(pwd, 'output', 'faster_rcnn_final', 'faster_rcnn_VOC0712_vgg_16layers'); %% VGG-16
model_dir                   = fullfile(pwd, 'output', 'faster_rcnn_final', 'faster_rcnn_VOC0712_ZF'); %% ZF

and then in $FRCN_ROOT run 'script_faster_rcnn_demo.m' you will see this

fast_rcnn startup done
001763.jpg (500x375): time 2.220s (resize+conv+proposal: 1.790s, nms+regionwise: 0.429s)
004545.jpg (500x375): time 2.727s (resize+conv+proposal: 1.962s, nms+regionwise: 0.765s)
000542.jpg (500x375): time 2.114s (resize+conv+proposal: 1.820s, nms+regionwise: 0.294s)
000456.jpg (500x375): time 2.461s (resize+conv+proposal: 1.822s, nms+regionwise: 0.639s)
001150.jpg (500x375): time 2.292s (resize+conv+proposal: 1.811s, nms+regionwise: 0.481s)
mean time: 2.363s
Cleared 0 solvers and 2 stand-alone nets

[jacobunderlinebenseal]

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

[sallymmx]

I have the same problem: Invalid MEX-file 'C:\Users\Public\faster_rcnn-master\external\caffe\matlab\caffe_fasterrcnn\caffe.mexw64': The specified

The simplest way to solve this problem is to check your environment variables. Check the cuda version in your "Path" have to be v-6.5. After I change my cuda version in "Path" from 7.5 to 6.5, the problem is solved and I can run the demo. Hope this can help, thank you!

[pallavimitra]

I have the same problem. I am using VS2013 and Cuda 8 in window 10. the error is like - fast_rcnn startup done GPU 1: free memory 7087721676 Use GPU 1 Invalid MEX-file 'D:\PALLAVI\faster_rcnn-master\faster_rcnn-master\external\caffe\matlab\caffe_fasterrcnn\caffe.mexw64': Missing dependent shared libraries: 'cublas64_65.dll' required by 'D:\PALLAVI\faster_rcnn-master\faster_rcnn-master\external\caffe\matlab\caffe_fasterrcnn\caffe.mexw64' 'curand64_65.dll' required by 'D:\PALLAVI\faster_rcnn-master\faster_rcnn-master\external\caffe\matlab\caffe_fasterrcnn\caffe.mexw64'.

Error in caffe.setdevice (line 9) caffe('set_device', device_id);

Error in active_caffe_mex (line 26) caffe.set_device(gpu_id-1);

Error in script_faster_rcnn_demo (line 10) active_caffe_mex(opts.gpu_id, opts.caffe_version);

Can anyone please help me?

[Robin970822]

Apparently, we almost have the same problem

[aous72]

I know this is an old thread, but I faced the same problem.
The problem is that MATLAB cannot find the dlls needed by caffe_.mexw64.

The solution is the add the folder with the necessary DLLs to the PATH environment variable; other solutions might also possible. For me, I created a startup.m inside C:\Users\YourUsername\Documents\MATLAB\startup.m and inside that file I put setenv('PATH', [getenv('PATH') 'C:\Users\YourUsername\Documents\MATLAB+caffe\private']); because all my dlls, including cudann, opencv, ... etc. are inside the private folder.

Hope this helps.

[saudpp]

@mindcont can you send me the link of caffe you installed