# ASVspoof2017 Database Interface¶

## ASVspoof2017 Protocols¶

ASVspoof2017 database provides one protocols:

* competition - protocol used in Automatic Speaker Verification Spoofing and Countermeasures Challenge (ASVspoof
2017). It includes training and development sets with labels and an anonymized evaluation set.


The protocol is supported by bob.db.asvspoof2017 DB interface for Bob.

## Getting the data¶

The original data and the description of protocols can be downloaded directly from ASVspoof2017.

## Using this interface¶

Once the interface package is installed, SQL database file need to be downloaded using the following command:

$bob_dbmanage.py asvspoof2017 download  This interface can be used to directly query and access the database protocols and samples, or/and in verification bob.bio. and PAD bob.pad. frameworks of Bob toolkit. The database filelist can be queried via the following command line: $ bob_dbmanage.py asvspoof2017 dumplist --help


To use the database in verification experiments within bob.bio. framework, a bob.bio.database entry point need to be defined in the setup.py file of the package that would run these experiments as so, as follows:

'bob.bio.database': [
'asvspoof2017             = bob.path.to.config.file:database',
]


To use the database in experiment within bob.pad. framework, a bob.pad.database entry point need to be defined in the setup.py file of the package that would run these experiments as so, as follows:

'bob.pad.database': [
'asvspoof2017             = bob.path.to.config.file:database',
]


The config file (other ways to defined the database are also available in Bob, please see database API documentation) would then initialize the database with the path to the directory where the actual database sample are located, see the following example for a bob.bio. package:

import bob.bio.base.database
asvspoof2017_input_dir = "PATH_TO_DATA"
database = bob.bio.base.database.ASVspoof2017BioDatabase(
protocol = 'competition-licit',
original_directory=asvspoof2017_input_dir,
original_extension=".wav",
training_depends_on_protocol=True,
)