Skip to content

Latest commit

 

History

History
153 lines (132 loc) · 4.75 KB

setupnotes.md

File metadata and controls

153 lines (132 loc) · 4.75 KB

Setting up development environment

If you wish to use an Nvidia GPU, go to the buttom of the guide and begin with the CUDA guide. Otherwise start from the top and ignore the CUDA instructions.

General tools and dependencies

Install and update essential system tools

sudo apt-get update && sudo apt-get upgrade
sudo apt-get install build-essential cmake cmake-gui g++ gfortran git pkg-config software-properties-common wget libatlas-dev libatlas-base-dev
sudo apt-get install python-dev python2.7-dev python3.5-dev python-pip python-numpy python-scipy python-nose python-h5py python-matplotlib python-pandas python-sympy python-pygments python-sphinx python-setuptools python-cvxopt

In some cases, the python tools are not availble in newest versions from apt-get, then use pip instead. An exampel would be: python-skimage. If an outdated version of python-skimage was installed using apt-get, just use pip to get a more recent version:

sudo pip uninstall scikit-image
sudo pip install scikit-image

Python

Installing python packages using pip

sudo pip install --upgrade pip
sudo pip install pillow
sudo pip install scikit-image
sudo pip install scikit-learn
sudo pip install cython
sudo pip install pystruct
sudo pip install pyqpbo

OpenCV

Get OpenCVs remaining dependencies ref

sudo apt-get install libjpeg8-dev libtiff5-dev libjasper-dev libpng12-dev libavcodec-dev libavformat-dev libswscale-dev libv4l-dev libxvidcore-dev libx264-dev libgtk-3-dev

Build OpenCV from source along with the contrib module

git clone https://github.com/opencv/opencv.git
git clone https://github.com/opencv/opencv_contrib.git
cd opencv/
mkdir release
cd release/
cmake -D OPENCV_EXTRA_MODULES_PATH=../../opencv_contrib/modules -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local ..
sudo make -j7 install

PCL

Get PCLs remaining dependencies

sudo apt-get install libeigen3-dev libvtk5-dev libusb-dev libgtest-dev git-core freeglut3-dev libxmu-dev libxi-dev libusb-1.0-0-dev graphviz mono-complete qt-sdk mpi-default-dev openmpi-bin openmpi-common libflann-dev libboost-all-dev

Build PCL from source

git clone https://github.com/PointCloudLibrary/pcl.git
cd pcl/
mkdir release
cd release/
cmake -DCMAKE_BUILD_TYPE=Release -DBUILD_GPU=ON -DBUILD_apps=ON -DBUILD_examples=ON ..
sudo make -j7 install

Tensorflow

Begin by setting a TF environment variable according to your setup: tensorflow

# Ubuntu/Linux 64-bit, GPU enabled, Python 2.7
# Requires CUDA toolkit 8.0 and CuDNN v5. For other versions, see "Installing from sources" below.
export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-0.12.0rc0-cp27-none-linux_x86_64.whl

Then install using pip

sudo pip install --upgrade $TF_BINARY_URL

Theano

sudo pip install Theano

(note) Switch between Theano using GPU or CPU by setting enviroment variable:

Python example:
os.environ["THEANO_FLAGS"] = "mode=FAST_RUN,device=gpu,floatX=float32"

(note) Increase performance slightly by reserving memory: add the following lines to .theanorc in home dir

[lib]
cnmem=.75

Keras

ref

sudo pip install keras

Configure keras to use Theano backend instead of Tensorflow

nano ~/.keras/keras.json
{
    "image_dim_ordering": "th",
    "epsilon": 1e-07,
    "floatx": "float32",
    "backend": "theano"
}

Installing cuda 8.0 Ubuntu

ref

CUDA

Download CUDA 8.0 .deb file, for your system, from nvidia-tools

sudo dpkg -i cuda-repo-ubuntu1604_8.0.44-1_amd64.deb
sudo apt-get update
sudo apt-get install cuda

Setting CUDA environment variables

echo 'export PATH=/usr/local/cuda/bin:$PATH' >> ~/.bashrc
echo 'export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc

Restart

sudo shutdown -r now

cuDNN

Login and download the cuDNN library that match your CUDA version through nvidias developer portal nvidia-dev

cd ~/Downloads/
tar xvf cudnn*.tgz
cd cuda
sudo cp */*.h /usr/local/cuda/include/
sudo cp */libcudnn* /usr/local/cuda/lib64/
sudo chmod a+r /usr/local/cuda/lib64/libcudnn*

Nvidia driver

NB: I recommend not installing the nvidia driver. Installing CUDA should take care of all you need. Here is how is can be done anyway.

sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt-get update
sudo apt-get install nvidia-367