cpp

EmotiEffLib C++ examples

Here you can find examples of using EmotiEffLib in C++.

Here is a list of examples:

• One image emotion recognition.ipynb describes how to use EmotiEffLib to recognize facial emotions on images.
• Predict emotions on video.ipynb describes how to use EmotiEffLib for predicting facial emotions on videos.
• Predict engagement and emotions on video.ipynb describes how to use EmotiEffLib for predicting facial expressions and recognizing a person's engagement in a video.

Building and running examples

To run the examples locally you need to do the following:

1. Install python dependencies:

pip install -r requirements.txt

2. Build EmotiEffCppLib with Libtorch and ONNXRuntime. It is important to build EmotiEffCppLib with flag: -DBUILD_TESTS=ON because we need to reuse one library which builds for tests. Also, it is required to build EmotiEffCppLib with flag: -DBUILD_SHARED_LIBS=ON because xeus-cling works only with shared libraries.
3. Install xeus-cling. Instruction how to build xeus-cling can be found here.

After installing xeus-cling, you should be able to check available kernels and see xcpp17 kernel:

$ jupyter kernelspec list
Available kernels:
  python3    /opt/anaconda3/envs/emotiefflib/share/jupyter/kernels/python3
  xcpp11     /opt/anaconda3/envs/emotiefflib/share/jupyter/kernels/xcpp11
  xcpp14     /opt/anaconda3/envs/emotiefflib/share/jupyter/kernels/xcpp14
  xcpp17     /opt/anaconda3/envs/emotiefflib/share/jupyter/kernels/xcpp17

4. Prepare models for cpp runtime:

python3 <EmotiEffLib_root>/models/prepare_models_for_emotieffcpplib.py

5. Download and unpack test data:

cd <EmotiEffLib_root>/tests
./download_test_data.sh
tar -xzf data.tar.gz

6. Run jupyter notebook and select C++ kernel.