Spear - Adding Custom Database

[Neve Stagy]

Hello,

I'm having trouble making sense of the steps needed to run experiments on a custom database. I've searched both the site and this group without much luck. I am attempting to use a portion of the Voxforge corpus partitioned as follows:

name: research_test_initial:

• Dev models: 40 samples
  
• Dev probes: 25 samples
  
• Eval models: 40 samples
  
• Eval probes; 25 samples
  
• UBM: 90 samples
  

Can anyone explain the next steps I'm missing? 

Thank you in advance!

[Neve Stagy]

To clarify, I am not sure how to "register" my database - running resources.py does not display it.

[Manuel Günther]

As long as you have your configuration file for the database, you don't need to register it. Please read: http://pythonhosted.org/bob.bio.base/filelist-guide.html and http://pythonhosted.org/bob.bio.base/more.html and tell us exactly, where you have problems understanding the pipeline.

Manuel

[Neve Stagy]

Sorry for not clarifying. I'm confused on the steps under Protocol and File Lists. I attempted to replicate how Voxforge was configured but can't identify where I need to place the necessary .py files. 

I always get the following error:

ImportError: The given command line option 'test' is neither a resource for a 'database', nor an existing configuration file, nor could be interpreted as a command (error: name 'test' is not defined)

[Manuel Günther]

You can place your database_config.py (this name is just an example, it can be called however you want) file anywhere you want, and then say:

verify.py -d <path_to_your_file>/database_config.py ...

The script will automatically detect that what you have passed is an actual file and use this file as a configuration file. You can mix-and-match configuration files as well as registered resources as you want, see: http://pythonhosted.org/bob.bio.base/more.html

[Neve Stagy]

I think I got it. Thank you!