User’s Guide

The FRGC (Face Recognition Grand Challenge) ver2.0 Database is provided by the NIST webpage: http://face.nist.gov/frgc/ Please contact the responsible person to get your own copy of the database (be aware that it is huge). Due to the size of the database, experiments using this package might require a lot of memory. Already running the tests will need up to 2 GB of free memory.

In opposition to other databases, there is no .sql3-file for this database, but instead the XML lists provided in the database are used directly. In order for the database interface to work properly, you have to specify the correct path on each usage. To avoid that, you can set the path in the bob.db.frgc.Interface.frgc_database_directory() function of the bob/db/frgc/driver.py file to your FRGC image database main directory. For use at Idiap, the right directory is preset.

In opposition to the original FRGC protocols, here only those image files and models that are required by the mask are used. This saves some time and space, but ensures identical results.

Warning

Do not store the model ids between sessions. These model id’s are generated on the fly and might change between database sessions.

Note

During one database session, model ids are unique and stable.

The Database Interface

The bob.db.frgc.Database complies with the standard biometric verification database as described in Common Functionality in all Verification Databases, implementing the interface bob.db.verification.utils.Database.

Todo

Explain the particularities of the bob.db.frgc.Database.