Installing the Intel® Computer Vision SDK on Linux*

Installing the Intel® Computer Vision SDK on Linux*

The installation package comes as an archive that contains components and installation scripts.

These instructions describe:

• Installation prerequisitess
• Pre-installation steps
• Two installation options: Using a graphical user interface or a command-line script
• Caffe* Installation

Prerequisites

The following are required to install all components and enable the full set of the Intel® Computer Vision SDK (Intel® CV SDK) features:

• CMake* version 2.8 or higher
• GCC* version 4.8 or higher
• OpenCV version 3.4
• Python* version 3.5 or higher

In addition, install the Caffe framework if you will use advanced functions of the Deep Learning Deployment Toolkit Model Optimizer. 

The steps in the next section direct you to use a bash script to install:

• Dependencies for Intel-optimized OpenCV 3.4, the Inference Engine, and Model Optimizer tools
• The OpenCL™ 2.0 GPU/CPU driver package for Linux*, required for applications that offload computation to your Intel® GPU.

Pre-installation Steps

Use these steps to prepare your development machine for the Intel CV SDK software.

1. Create a new directory. this example uses a directory called install:

   mkdir ~/install

2. Go to the new directory. Move the downloaded archive package to this directory and unpack the archive. This example uses the Downloads directory:

   cd ~/install
   mv ~/Downloads/intel_cv_sdk_<version>.tgz
   tar -xavf intel_cv_sdk_<version>.tgz

3. Go to the directory with the unpacked files. This is intel_cv_sdk <version>

   cd intel_cv_sdk_<version>

4. Run install_cv_sdk_dependencies.sh to install the external dependencies. These dependencies are the packages required for Intel-optimized OpenCV 3.4, the Inference Engine, and the Model Optimizer tools:

   ./install_cv_sdk_dependencies.sh

5. For applications that offload computation to your Intel® GPU, the OpenCL™ 2.0 GPU/CPU driver package for Linux* is required. To install the driver, run the install_OCL_driver.sh script from the installation package directory. 

Using the GUI Installation Wizard

1. After you complete the pre-installation steps, run install_GUI.sh to start the GUI-based installation wizard:

   ./install_GUI.sh

2. Select an elevation mode, and then follow the on-screen instructions:
3. The Prerequisites screen tells you if you are missing any required or recommended components, and the effect the missing component has on installing or using the product. If you are missing a critical component, resolve the issue and then click Re-check.
4. When all critical issues are resolved, you will be able to click Next to begin the installation, or make final changes to your component selections and choose your installation directory.
5. The Installation summary screen shows you the options that will be installed if you make no changes.
6. If you want to change the selected components and/or specify an installation destination directory, click Customize:
7. A Complete screen indicates the software is installed. Click Finish to close the wizard and open the Getting Started page, which contains information to get you started quickly with Intel CV SDK.

Installing from an Installation Script

If you want to install the Intel CV SDK from a command line after you're done with the pre-installation steps, run ./install.sh script and follow the on-screen instructions.

The install.sh options are:

-h, --help - display help message

-v, --version - display version information

-s, --silent [FILE] - run setup silently using settings in the specified configuration file instead of from screen prompts

-d, --duplicate [FILE] - run setup interactively, and record the user input in the configuration file

-l, --lang - set user interface language

-t, --tmp-dir [DIRECTORY] - specify a temporary directory

-D, --download-dir [DIRECTORY] - specify a download directory

--user-mode - run setup with the current user privileges

--ignore-signature - skip signature validation

--ignore-cpu - skip the CPU model check

--nonrpm-db-dir [DIRECTORY] - set a directory for the installation database

--SHARED_INSTALL - install to a network-mounted drive or shared file system. This is typically for multiple users

During the installation you can:

• Choose the elevation mode for installation:
  • root
  • sudo
  • current user
• Chose the installation directory:
  • Default directory for root user is /opt/intel/computer_vision_sdk_<version>
  • Default directory for non-root user is ~/intel/computer_vision_sdk_<version>
• Customize components to install.

Your installation is complete. Continue with Getting Started to help you get you started quickly with the Intel CV SDK.

What You Installed

The installation created this directory structure:

In addition to these directories, a symlink to the installation directory is added to the /opt/intel/ directory. If you install multiple versions of the Intel CV SDK, the versions are installed side-by-side and the symlink /opt/intel/computer_vision_sdk points to the latest version.

Installing the Caffe* Framework

The Caffe* deep learning framework is required if your deployed Caffe topology has layers that are not implemented in the Model Optimizer and you do not register them as "Custom Operations".

To build a Caffe* framework with Python 3.5:

1. Use the commands:

   export CAFFE_HOME=PATH_TO_CAFFE
   cd $CAFFE_HOME
   rm -rf  ./build
   mkdir ./build
   cd ./build
   cmake -DCPU_ONLY=ON -DOpenCV_DIR=<your_opencv_install_dir> -DPYTHON_EXECUTABLE=/usr/bin/python3.5 ..
   make all # also builds pycaffe
   make install
   make runtest # optional

2.  Add the Caffe Python directory to PYTHONPATH to import it from the Python program:

   export PYTHONPATH=$CAFFE_HOME/python;$PYTHONPATH

3. Check the Caffe installation:

   python3 import caffe

If Caffe is installed correctly, the Caffe module is imported without errors.

Next Steps

Your installation is complete. Return to the Getting Started page for information to get started quickly.