Permalink
Switch branches/tags
Nothing to show
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
62 lines (50 sloc) 1.85 KB
------------------
Dependencies
-----------------
- CUDA (>= 4.0) (Not required for the CPU-only feature extractor).
- HDF5 (>= 1.8.11)
Download the source code http://www.hdfgroup.org/ftp/HDF5/prev-releases/hdf5-1.8.12/src/hdf5-1.8.12.tar.gz
To install locally -
$ ./configure --prefix=~/local/ --enable-threadsafe --with-pthread=/usr/include
$ make
$ make install
- Protocol Buffers (>= 2.5.0)
Download the source code https://protobuf.googlecode.com/files/protobuf-2.5.0.tar.gz
To install locally -
$ ./configure --prefix=~/local/
$ make
$ make install
- OpenCV
Download the source code https://github.com/Itseez/opencv/archive/3.0.0-alpha.zip
To quick install -
$ cmake-gui, set CMAKE_BUILD_TYPE to Release
$ cd build directory
$ make -j7
$ sudo make install
- libjpeg
This is often present on standard operating systems since it is used by a lot of programs.
It can be downloaded from http://libjpeg.sourceforge.net/
- This code uses C++11. Some features require gcc >= 4.6.
---------------------------
Hardware requirements
---------------------------
NVIDIA GPU(s).
Multi-GPU feature requires Fermi boards.
-------------------
Compilation
-------------------
Compiling the CPU-only feature extractor -
- Set the paths to the dependencies in `convnet/cpu/Makefile`.
- Run make in `convnet/cpu`.
Compiling the full package -
- Compile cudamat : Set the path to dependencies in `convnet/cudamat/Makefile` and run make in `convnet/cudamat/`.
- Set the paths to the dependencies in `convnet/Makefile`.
- Set the compute capability for your GPU.
- Run make in `convnet`.
Add convnet/bin to your PATH variable (maybe in your .bashrc)
---------------------
Run test examples
---------------------
Examples of nets on MNIST and ImageNet are included in
examples/mnist and examples/imagenet.
Follow the instructions in the respective README.md files.