The MATLAB program in Caffe supports 4. 7 of GCC and UBUNTU14. 04 of the band's own GCC is 4. 8 so it will be wrong to compile. So we'll install GCC4 first. 7, installation method can be checked online, as if sudo apt-get install gcc-4.7 and there are two versions of GCC, so you need to set the default GCC version in the following way we will install the g++ also installed on the g++4.7, so we replace the following: CD /usr/binsudo mv gcc gcc.baksudo ln-s gcc-4.7 gccsudo mv g++ g++.baksudo ln-s g++-4.7 g++ It's all right then Caffe from start to finish again, this compilation The final use of sudo make all Matcaffe for the compilation of MATLAB and make mattest test whether the installation is successful, but I do mattest and do not respond, detailed can refer to HTTP. Caffe.berkeleyvision.org/tutorial/interfaces.html website about the use of MATLAB interface compiled well after the Mex file, which is similar to the C + + project DLL file, is matlab Dynamic link library (so to understand), enter MATLAB2013A,CD to Caffe/matlab/caffe, input im = Imread (' peppers.png ');
[Scores, Maxlabel] = Matcaffe_demo (im,1);
If the error, may be in the Examples/imagent no training set file Caffe_reference_imagenet_model, the web has related files, no longer say more. As for the Python interface compilation I think similar to this principle, using this project to change the path, may eventually generate DLL to PYD and then import Caffe. has not yet been tried, providing only this
Using the Matlab interface in Caffe