File format

FaceGuard saves its database as two files:

• DatabaseName.facedb
• DatabaseName.facedb.nlabels

The .facedb file contains the actual face recognizer as serialized by OpenCV. It contains the information acquired during the training, which means the original input files are not needed. However this format lacks any other parameters, like number of components, etc. Unfortunately, OpenCV is not capable to use strings as labels for faces, so another file is required.

The .nlabels file is a JSON document that contains the actual names of the users along with their internal ID or label in our terminology.

The authors acknowledge that it would be desirable to combine these files, however OpenCV's limitation prevents this, since it only accepts a file path as an input parameter for serialization and deserialization.