occlum/demos/openvino/README.md
LI Qing b9fa937504 Add Intel OpenVINO demo
This demo shows how to run Intel OpenVINO Inference Engine on Occlum inside SGX enclaves
2019-12-10 06:07:27 +00:00

1.0 KiB

Use OpenVINO Inference Engine in SGX with Occlum

This project demonstrates how Occlum enables OpenVINO in SGX enclaves.

Step 1: Install CMake(3.15.5), because OpenVINO depends on a newer version of CMake

./install_cmake.sh

Step 2: Download OpenVINO and build the Inference Engine, it will also download and install OpenCV

./download_and_build_openvino.sh

When completed, the resulting OpenVINO can be found in openvino_src directory. Threading Building Blocks (TBB) is used by default. To use OpenMP, add option --threading OMP when invoking the script above.

Step 3: Download the example of OpenVINO models from 01.org

./download_openvino_model.sh

Step 4: Run OpenVINO Inference Engine benchmark app inside SGX enclave with Occlum

./run_benchmark_on_occlum.sh

Step 5 (Optional): Run OpenVINO Inference Engine benchmark app in Linux

./openvino_src/inference-engine/bin/intel64/Release/benchmark_app -m ./model/age-gender-recognition-retail-0013.xml