FaceRecogClientCpp

Requirements

Tested on Ubuntu 16.04 and 17.04

Please see FaceRecogClientCppDocker for an automated build enviroment.

You need to build the project and run cpack to create the debs.

apt-get install libopencv-dev, qtbase5-dev, libqt5webkit5-dev, libopenblas-dev, liblapack-dev
dpkg -i face-recog-client.deb

The embedding service Face Recog Python Services must run and the tracking service Face Recog Server must be available.

The /usr/local/etc/face-recog-config.ini config file needs to be updated:

embedding_service_url = "http://localhost:5001/"
- Preferably on the same machine for better performance and network delay.
tracking_api_url = "https://face.otep.ch/"
- The single endpoint for all camera instances to send camera images and embeddings to.
location_name = "cam2"
- Should be unique across all camera instances.
tracking_view_url = "https://face.otep.ch/locations/$location_name/current"
- Website with recognized faces. Should be the same server that handles the tracking api requests.
predictor_path = "${FACE_RECOG_LANDMARK_PATH}"
- Landmark file with the predictor shapes for facial detection.

Faces are detected with dlib's frontal face detector (http://dlib.net/) Face embeddings are calculated using the Tensorflow-based facenet (https://github.com/davidsandberg/facenet)