Install Caffe With Anaconda

Tuesday, August 30, 2016

If you want to install Caffe on Ubuntu 16.04 along with Anaconda, here is an installation guide:

Install Nvidia driver and Cuda (Optional)

If you want to use GPU to accelerate, follow instructions here to install Nvidia drivers, CUDA 8RC and cuDNN 5 (skip caffe installation there).

Install Anaconda

Download Anaconda from here. Choose Python 2.7 version 64-BIT INSTALLER to install it. Then update it:

conda update conda

If you want to create an environment such as testcaffe, execute commands:

conda create -n testcaffe python
source activate testcaffe

Install OpenCV:

conda install -c menpo opencv3

Install Dependencies

sudo apt-get update

sudo apt-get upgrade

sudo apt-get install -y build-essential cmake git pkg-config

sudo apt-get install -y libprotobuf-dev libleveldb-dev libsnappy-dev protobuf-compiler

sudo apt-get install -y libatlas-base-dev 

sudo apt-get install -y --no-install-recommends libboost-all-dev

sudo apt-get install -y libgflags-dev libgoogle-glog-dev liblmdb-dev

Build Caffe

Go to, download zip archive and unpack it. Or clone the source code. Enter the <caffe-home> directory in the terminal window:

mkdir build
cd build
cmake ..
make all
make install

Enter <caffe-home>/python directory to install Python packages:

conda install cython scikit-image ipython h5py nose pandas protobuf pyyaml jupyter
for req in $(cat requirements.txt); do pip install $req; done
cd ../build
make runtest

If you want to use some other packages in the Conda evnviroment, you need to install them now, otherwise the packages might not find the dependences you installed in the evnviroment.

Add the module directory to your $PYTHONPATH by

cd ../python

Test Run

First verify the installation:

python -c "import caffe;print caffe.__version__"

If the Caffe version number is shown up correctly, then change directory to examples, execute command:

jupyter notebook

Choose one of the notebook examples to test the Caffe installation.