private dataset

[hanan.sh...@gong.io]

New to bob-spear.
I am trying to figure out how should one use the package to train a ubm, enroll and test with proprietary wav files not of a public dataset.
In other words, do bob.spear has documentation on how to setup a new dataset?
Best,
Hanan

[Pavel Korshunov]

Hi,

You can create file lists for your data in a specific format and setup a new database interface in a few lines. Please take a look here:

http://pythonhosted.org/bob.bio.base/implementation.html#verification-database-interface

and here:

http://pythonhosted.org/bob.bio.base/filelist-guide.html

For example, once you create file lists, say inside folder 'folder_filelists', you can define a database for your experiments like so:

import bob.bio.base

from bob.bio.spear.database import AudioBioFile

class MyBeautyDB(bob.bio.base.database.FileListBioDatabase):

  """Wrapper class for MyBeautyDB speaker database.

  """

  def __init__(self, original_directory="[MyBeautyDB_DATA_DIRECTORY]", original_extension=".wav"):

    # call base class constructor

    from pkg_resources import resource_filename

    folder = resource_filename(__name__, '../folder_filelists')

    super(MyBeautyDB, self).__init__(folder, 'db_name', bio_file_class=AudioBioFile,

                                   original_directory=original_directory,

                                   original_extension=original_extension)

database = MyBeautyDB()

Now, you can use 'database' object inside your configuration scripts that you can pass to verify.py or train.py (or train_gmm.py, train_isv.py, etc) of bob.bio.spear.

I hope it helps.

Pavel

--
-- You received this message because you are subscribed to the Google Groups bob-devel group. To post to this group, send email to bob-...@googlegroups.com. To unsubscribe from this group, send email to bob-devel+unsubscribe@googlegroups.com. For more options, visit this group at https://groups.google.com/d/forum/bob-devel or directly the project website at http://idiap.github.com/bob/
---
You received this message because you are subscribed to the Google Groups "bob-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bob-devel+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Dr. Pavel Korshunov

Biometric group

Idiap Research Institute

Rue Marconi 19

CH - 1920 Martigny

Switzerland

Room: 207

Tel:  +41 27 721 77 56

Fax: +41 27 721 77 12

[Hanan Shteingart]

Thanks,

Can you please give a more detailed example?

Say I have n-files wav files with m(n) segments in each file (start&end in seconds and speaker id) - how would one enter this information within the dataset?

Best,

Hanan

You received this message because you are subscribed to a topic in the Google Groups "bob-devel" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/bob-devel/mDSXKoaFvGA/unsubscribe.
To unsubscribe from this group and all its topics, send an email to bob-devel+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

Dr. Hanan Shteingart | Data Scientist | Gong.io
P: +972-54-2271572
M: hanan.sh...@gong.io
W: www.gong.io

Love Gong? Share the love and help us grow.

[Manuel Günther]

Hi Hanan,

I am afraid that it is currently not possible to have samples from different speakers in the same file. The only solution I can think of is to split the data into several files with an external application and store the relevant parts of it separate files, one per speaker and sample.

Best regards
Manuel

[Amir Mohammadi]

Hi,

It is possible to have a database interface that handles your files but maybe per-processing your data as Manuel explained is easier.

In order to do so, you need to create a database interface and not use the filelist database interface. This is best explained in bob.db.base docs:https://www.idiap.ch/software/bob/docs/bob/bob.db.base/master/ especially the development guide and then in https://www.idiap.ch/software/bob/docs/bob/bob.bio.base/master/implementation.html#databases .
An example is also the bob.db.nist_sre12 database where each file contains two samples: https://gitlab.idiap.ch/bob/bob.db.nist_sre12/blob/14dd6ce6b3bc3ec1cce8b5eb858c0c21760dab4a/bob/db/nist_sre12/models.py#L124  and its high-level interface in https://gitlab.idiap.ch/bob/bob.bio.spear/tree/add_nistsre12_db_

but again it's easier to pre-process your files and create files per sample, per speaker.

Best,

Amir

--
-- You received this message because you are subscribed to the Google Groups bob-devel group. To post to this group, send email to bob-...@googlegroups.com. To unsubscribe from this group, send email to bob-devel+...@googlegroups.com. For more options, visit this group at https://groups.google.com/d/forum/bob-devel or directly the project website at http://idiap.github.com/bob/

---
You received this message because you are subscribed to the Google Groups "bob-devel" group.

To unsubscribe from this group and stop receiving emails from it, send an email to bob-devel+...@googlegroups.com.