30 lines
1.0 KiB
Markdown
30 lines
1.0 KiB
Markdown
# Use OpenVINO Inference Engine in SGX with Occlum
|
|
|
|
This project demonstrates how Occlum enables [OpenVINO](https://github.com/opencv/dldt) 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](https://download.01.org/opencv/)
|
|
```
|
|
./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
|
|
```
|