Real-time face landmarks detection using face-alignment library and opencv

In this Project we used face-alignment library and opencv to detect face landmarks on a realtime/offline video. One can use either camera (realtime) or an input video (offline) to do the inference.

How to run

Instructions to run in linux bash:

• Having the environment ready:
  • Using conda
    • conda create -n rtfd_env python=3.9
    • conda activate rtfd_env
  • Using pip
    • virtualenv -p python3.9 rtfd_env
    • source rtfd_env/bin/activate
• Install dependencies:
  • Using conda
    • conda install pytorch torchvision cudatoolkit=11.3 -c pytorch (See pytorch installation guide)
    • conda install -c 1adrianb face_alignment (see face-alignment Github repo)
  • Using pip
    • pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu113
    • pip install face_alignment
• Run:
  • Inference from camera: python main.py
  • Inference from an input video: python main.py ./input/v1.MP4

(Alternatively, one can use run.sh as a stand-alone script to do the above)