Celantur SDK Examples

The Celantur SDK provides privacy-preserving image blurring through automatic anonymisation of faces and license plates.

In this repository you find a lot of example code how to use Celantur SDK.

Prerequisite

The SDK consists of several shared objects and include files, that can be linked against and included in your program. It is shipped as a Debian package and the files are extracted to /usr/local/.

See Celantur Doc for installation instructions.

Building examples

The easiest way to compile the examples is using CMake.

First, create CMake configuration files in the repo folder:

mkdir -p build/
# Create configuration files
cmake -S . -B build/
# Build the binaries
cd build/
make

In CMakeLists.txt you can use the findPackage directive to add libraries and link targets into your program, e.g.:

find_package(CppProcessing REQUIRED CelanturSDK common-module)
...
target_link_libraries(YourExecutableOrLibrary PRIVATE CppProcessing::CelanturSDK CppProcessing::common-module)

If CMake findPackage returns an error, check out the Troubleshooting section.

Run examples

Assets

Copy all assets required to assets/:

• license: License file.
• v6-static-fp32.onnx.enc: Encrypted model.
• image.jpg: Test images.

In case you were provided with some custom models, such as less-precise but faster models, your model name can be different.

You can change the file names in the source code.

Executables

In build/, you find the following executables:

• quickstart: Blurs the image.
• processor_parameters_detail: Same as quickstart but with more parameters (2x2 tiling and blurring of only the top-left quarter of the image).
• decoding_encoding: Image is encoded and decoded with the SDK (instead of OpenCV), and the metadata is conserved in the output image.
• inference_tinkering_onnx: The code shows how to configure the parameters of the ONNX inference engine.
• inference_tinkering_openvino: The code shows how to configure the parameters of the OpenVINO inference engine.
• detections_and_thresholds: The code shows how to filter detections based on threshold, to visualise detections and to generate detection metrics in JSON.
• person-vehicle-anon.cpp: The code shows how to anonymise persons and vehicles with segmentation mask.
• small_model.cpp: The code shows how to use a smaller and faster model for anonymisation on the example of OpenVINO inference engine.
• detect.cpp: How to use SDK to only get detections and don't spend any time on actual blurring.

Troubleshooting

CppProcessing is not found

CMake Error at CMakeLists.txt:10 (find_package):
  By not providing "FindCppProcessing.cmake" in CMAKE_MODULE_PATH this
  project has asked CMake to find a package configuration file provided by
  "CppProcessing", but CMake did not find one.

  Could not find a package configuration file provided by "CppProcessing"
  with any of the following names:

    CppProcessingConfig.cmake
    cppprocessing-config.cmake

  Add the installation prefix of "CppProcessing" to CMAKE_PREFIX_PATH or set
  "CppProcessing_DIR" to a directory containing one of the above files.  If
  "CppProcessing" provides a separate development package or SDK, be sure it
  has been installed.

Solution

Add -DCppProcessing_DIR=/usr/local/lib/cmake to the cmake configuration.

cmake -S . -B build/ -DCppProcessing_DIR=/usr/local/lib/cmake

Error loading shared libraries

./celantur_sdk_example: error while loading shared libraries: libprocessing.so: cannot open shared object file: No such file or directory`

Solution

Add the library path to the linker paths:

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/lib/

Or edit the runpath.

Additional settings

By default, logging level is set to INFO. If you find it too verbose, you can change it to WARNING or ERROR. Here is an example how to set logging level to WARNING:

export LOG_LEVEL=WARNING