Intel's Deep Learning Inference Engine backend

Intel's Deep Learning Inference Engine is a part of Intel® Computer Vision SDK. You can use it as a computational backend for OpenCV deep learning module.

• Download and install Intel's Deep Learning Deployment Toolkit.

• Build OpenCV specifying path to installed libraries and plugins.

  • Ubuntu

  cmake \
    -DWITH_INF_ENGINE=ON \
    -DINTEL_CVSDK_DIR=/opt/intel/deeplearning_deploymenttoolkit/deployment_tools/ \
    -DIE_PLUGINS_PATH=/opt/intel/deeplearning_deploymenttoolkit/deployment_tools/inference_engine/lib/ubuntu_16.04/intel64/ \
    -DENABLE_CXX11=ON \
    ...
  

  • Microsoft Windows

  cmake ^
    -DWITH_INF_ENGINE=ON ^
    -DINTEL_CVSDK_DIR=C:\\Intel\\DeepLearning-DeploymentToolkit_1.0.5852 ^
    -DIE_PLUGINS_PATH=C:\\Intel\\DeepLearning-DeploymentToolkit_1.0.5852\\lib\\intel64\\Release ^
    -DENABLE_CXX11=ON ^
    ...
  

  Add path to Intel's Inference Engine plugins into PATH variable:

  set PATH=C:\Intel\DeepLearning-DeploymentToolkit_1.0.5852\bin\intel64\Release;%PATH% 
  

• Enable Intel's Inference Engine backend right after cv::dnn::readNetFrom* invocation:

  net.setPreferableBackend(DNN_BACKEND_INFERENCE_ENGINE);