##face_net_recognizer its a ros package depend on tensorflow and dlib tensorflow -- https://www.tensorflow.org/ dlib -- http://dlib.net/
the face_net part is based on https://github.com/davidsandberg/facenet
in <face_net_recognizer>\scripts\face_recognize\model_data
has two file
model.ckpt-500000 which is a tensorflow checkpoint file
The data has been pre-processed as described on the OpenFace web page (https://cmusatyalab.github.io/openface/models-and-accuracies/), i.e. using ./util/align-dlib.py data/lfw/raw align outerEyesAndNose data/lfw/dlib-affine-sz:96 --size 96 --fallbackLfw data/lfw/deepfunneled
the shape_predictor_68_face_landmarks.dat is a face_landmark file based on dlib
in the <face_net_recognizer>\scripts\face_recognize\templates we have two sample people and a describe_csv file
change them to what ever you want to use in the face_recognitio :)
facenet is based on google's paperFaceNet: A Unified Embedding for Face Recognition and Clustering it use a novel method mot formillar with the classfiermodel so only one templates and on training would perfomes well :)