Facereclib with BOB

[Mathew iprocessor]

Hi I have downloaded the Facereclib and followed the steps from this link https://pypi.python.org/pypi/facereclib/1.0.0a1.

But I am getting error as "Error: Couldn't find a distribution for 'bob==1.2.0'." after running the step "bin/buildout".

But I have installed BOB also. Still I am facing problem.

1) How to check whether bob is installed or not? I can see the ".so" files in /usr/local/lib

2) How to tests the algorithms installed in Facereclib.

I am using ubuntu machine,Detailed steps would be great.

[Laurent El Shafey]

Hello,

Which operating system are you using?
How did you install bob?
Why are you using an old release of the facereclib? (current version is 1.1.3)
https://pypi.python.org/pypi/facereclib/

1. To check if bob is installed properly:
Could you try to open a python interpreter, to run the following two commands:
>>> import bob
>>> print bob.version
And to tell us the output.

2. To test the algorithms provided by the Facereclib, just follow the instructions here:
https://pypi.python.org/pypi/facereclib

I'm wondering if the error message you're getting is just because you are using the facereclib with bob 1.2.1 installed.
The facereclib is currently designed to work against bob 1.2.0, whereas we just released bob 1.2.1.
A version of the facereclib compatible with bob 1.2.1 will be available very soon.

I guess that we could help you further once you will provide the information above.

Cheers,
Laurent

--
-- 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.
For more options, visit https://groups.google.com/groups/opt_out.

[Mathew iprocessor]

1) I am using ubuntu 12.04.

2) I downloaded and build the bob library and installed using "Make install"

3) I gave a wrong link. I am using facereclib version 1.1.3.

I have checked as you have mentioned "using import command". there are no errors.

You are right, The version of bob I have build and installed is 1.2.1.

Now I downloaded bob1.2.0 and tried building again but I am getting the below error.

[  7%] Building CXX object src/io/cxx/CMakeFiles/bob_io.dir/VideoUtilities.cc.o

/home/tsk/bob-1.2.0/src/io/cxx/VideoUtilities.cc:49:9: error: ‘CodecID’ does not name a type

make[2]: *** [src/io/cxx/CMakeFiles/bob_io.dir/VideoUtilities.cc.o] Error 1

make[1]: *** [src/io/cxx/CMakeFiles/bob_io.dir/all] Error 2

make: *** [all] Error 2

Can you please help me out.

[Laurent El Shafey]

I'm surprised by the compilation error you're getting. Have you installed all the required dependencies and does cmake initially reports any error?

I've just checked on my side, and downloading bob 1.2.0 from http://goo.gl/EJJC3, I was able to build it on an up-to date ubuntu 12.04 64 bits platform without any problem.

If you have root access to your machine, you could consider using directly our ubuntu ppa, as described here:
https://github.com/idiap/bob/wiki/Packages

which means:

$ sudo add-apt-repository ppa:biometrics/bob
$ sudo apt-get update
$ sudo apt-get install bob

This will install bob 1.2.1 that you already have somewhere.

To circumvent the problem you were initially getting when using bob 1.2.1 ("Error: Couldn't find a distribution for 'bob==1.2.0'." after running the step "bin/buildout"), you could temporarily update your setup.py file from the facereclib package as follows:
Update the line 136:
"bob == 1.2.0",
by
"bob >= 1.2.0",

Then the error message will disappear BUT some of the tests (ISV, JFA) from the test suite will fail because of a change in the random number generator (cf. issue #116, https://github.com/idiap/bob/issues/116). However, this won't prevent you to run any face recognition experiments.

Cheers,
Laurent

[Mathew iprocessor]

Thanks for your reply. I changed the setup file tp ">=1.2.0" and it got build successfully ( facereclib alone ).

My intention is to use and test the CSU-facerecognition algorithms. I followed the steps here (https://github.com/bioidiap/xfacereclib.paper.BeFIT2012) to make a build, but I got stuck with the same issue ( bob1.2.0 missing )

And below is the log for your reference. I am in no clue in which file I need to change now, I have checked the "setup.py" file and there was no command in "xfacereclib.extension.CSU-1.0.0" package.

The below is the log for your reference:

mr.developer: Queued 'PythonFaceEvaluation' for checkout.

mr.developer: Filesystem package 'PythonFaceEvaluation' doesn't need a checkout.

Develop: '/home/tsk/PythonFaceEvaluation20120807/PythonFaceEvaluation'

Develop: '/home/tsk/xfacereclib.extension.CSU-1.0.0/.'

Uninstalling scripts.

Installing _mr.developer.

Generated script '/home/tsk/xfacereclib.extension.CSU-1.0.0/bin/develop'.

Installing scripts.

Develop distribution: matplotlib 1.1.1rc

uses namespace packages but the distribution does not require distribute.

Getting distribution for 'facereclib'.

Got facereclib 1.1.3.

Getting distribution for 'xbob.db.atnt'.

Got xbob.db.atnt 1.1.1.

Getting distribution for 'bob==1.2.0'.

While:

  Installing scripts.

  Getting distribution for 'bob==1.2.0'.

Error: Couldn't find a distribution for 'bob==1.2.0'.

I think I am following the correct steps(link)? Please advice.

[Laurent El Shafey]

You have two options:
- Either you understand why you are not able to build bob 1.2.0.
- Otherwise, you have to deal with bob 1.2.1 and to disable in all the
packages the requirement of this specific release bob 1.2.0. Looking at
this package (xfacereclib.paper.BeFIT2012), I can see that it again
requires bob 1.2.0 (bob == 1.2.0), and you have to update this
accordingly. To sum up, if you are installing
xfacereclib.paper.BeFIT2012 with bob 1.2.1, you have to update the
setup.py of this package to be able to use bob 1.2.1 AND you have to
update the one from the facereclib dependency (likely in the
eggs/facereclib-...) subdirectory.

Things will be much easier when new versions of these packages will be
available for bob 1.2.1.

Cheers,
Laurent

On 2013-10-17 15:09, Mathew iprocessor wrote:
> Thanks for your reply. I changed the setup file tp ">=1.2.0" and it
> got build successfully ( facereclib alone ).
>
> My intention is to use and test the CSU-facerecognition algorithms. I
> followed the steps here

> (https://github.com/bioidiap/xfacereclib.paper.BeFIT2012 [10]) to

>> http://goo.gl/EJJC3 [7], I was able to build it on an up-to date

>> ubuntu 12.04 64 bits platform without any problem.
>>
>> If you have root access to your machine, you could consider using
>> directly our ubuntu ppa, as described here:

>> https://github.com/idiap/bob/wiki/Packages [8]

>>
>> which means:
>>
>> $ sudo add-apt-repository ppa:biometrics/bob
>> $ sudo apt-get update
>> $ sudo apt-get install bob
>> This will install bob 1.2.1 that you already have somewhere.
>>
>> To circumvent the problem you were initially getting when using bob
>> 1.2.1 ("Error: Couldn't find a distribution for 'bob==1.2.0'." after
>> running the step "bin/buildout"), you could temporarily update your
>> setup.py file from the facereclib package as follows:
>> Update the line 136:
>> "bob == 1.2.0",
>> by
>> "bob >= 1.2.0",
>>
>> Then the error message will disappear BUT some of the tests (ISV,
>> JFA) from the test suite will fail because of a change in the random
>> number generator (cf. issue #116,

>> https://github.com/idiap/bob/issues/116 [9]). However, this won't

>>>> https://pypi.python.org/pypi/facereclib/ [5]

>>>>
>>>> 1. To check if bob is installed properly:
>>>> Could you try to open a python interpreter, to run the
>>>> following two commands:
>>>> >>> import bob
>>>> >>> print bob.version
>>>> And to tell us the output.
>>>>
>>>> 2. To test the algorithms provided by the Facereclib, just
>>>> follow the instructions here:

>>>> https://pypi.python.org/pypi/facereclib [6]

>>>>
>>>> I'm wondering if the error message you're getting is just
>>>> because you are using the facereclib with bob 1.2.1 installed.
>>>> The facereclib is currently designed to work against bob 1.2.0,
>>>> whereas we just released bob 1.2.1.
>>>> A version of the facereclib compatible with bob 1.2.1 will be
>>>> available very soon.
>>>>
>>>> I guess that we could help you further once you will provide
>>>> the information above.
>>>>
>>>> Cheers,
>>>> Laurent
>>>>
>>>> On 17/10/13 10:59, Mathew iprocessor wrote:
>>>>
>>>>> Hi I have downloaded the Facereclib and followed the steps
>>>>> from this link https://pypi.python.org/pypi/facereclib/1.0.0a1

>>>>> [1].

>>>>>
>>>>> But I am getting error as "Error: Couldn't find a distribution
>>>>> for 'bob==1.2.0'." after running the step "bin/buildout".
>>>>>
>>>>> But I have installed BOB also. Still I am facing problem.
>>>>>
>>>>> 1) How to check whether bob is installed or not? I can see the
>>>>> ".so" files in /usr/local/lib
>>>>> 2) How to tests the algorithms installed in Facereclib.
>>>>>
>>>>> I am using ubuntu machine,Detailed steps would be great. --
>>>>> -- 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 [2] or directly
>>>>> the project website at http://idiap.github.com/bob/ [3]

>>>>> ---
>>>>> 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.
>>>>> For more options, visit

>>>>> https://groups.google.com/groups/opt_out [4].
>
>
> Links:
> ------
> [1] https://pypi.python.org/pypi/facereclib/1.0.0a1
> [2] https://groups.google.com/d/forum/bob-devel
> [3] http://idiap.github.com/bob/
> [4] https://groups.google.com/groups/opt_out
> [5] https://pypi.python.org/pypi/facereclib/1.0.0a1
> [6] https://pypi.python.org/pypi/facereclib
> [7] http://goo.gl/EJJC3
> [8] https://github.com/idiap/bob/wiki/Packages
> [9] https://github.com/idiap/bob/issues/116
> [10] https://github.com/bioidiap/xfacereclib.paper.BeFIT2012

[Mathew iprocessor]

Ya I have tried that and in "xfacereclib.paper.BeFIT2012" there is a "setup.py" where there is no word "bob==1.2.0" and in eggs/facereclib-1.1.3-py2.7.egg/facereclib/ I coundnt find a "setup.py" file at all.

Reg: building bob1.2.0 do I need to uninstall the previous version and try? Actually I have build 1.2.1 version successfully, any idea?

[Laurent El Shafey]

If you've downloaded the latest version (1.0.0) of
"xfacereclib.paper.BeFIT2012" from PyPI:
https://pypi.python.org/packages/source/x/xfacereclib.paper.BeFIT2012/xfacereclib.paper.BeFT2012-1.0.0.zip
there is definitely a "bob == 1.2.0" at line 65. You need to update
that:

If you want to install another version of bob, yes it is better to
uninstall the previous version that you have. You could also try to
install our old deb package of bob 1.2.0 for lucid, instead of compiling
from source:
https://launchpad.net/~biometrics/+archive/bob/+files/bob_1.2.0-0%7Eppa1%7Elucid1_amd64.deb
(for amd64)
https://launchpad.net/~biometrics/+archive/bob/+files/bob_1.2.0-0%7Eppa1%7Elucid1_i386.deb
(for i386)

I will check later the content of
eggs/facereclib-1.1.3-py2.7.egg/facereclib/. Just let me know if you
succeed in the meantime.

Cheers,
Laurent

>>> (https://github.com/bioidiap/xfacereclib.paper.BeFIT2012 [1] [10])

>>>> http://goo.gl/EJJC3 [2] [7], I was able to build it on an up-to

>> date
>>>> ubuntu 12.04 64 bits platform without any problem.
>>>>
>>>> If you have root access to your machine, you could consider using
>>
>>>> directly our ubuntu ppa, as described here:

>>>> https://github.com/idiap/bob/wiki/Packages [3] [8]

>>>>
>>>> which means:
>>>>
>>>> $ sudo add-apt-repository ppa:biometrics/bob
>>>> $ sudo apt-get update
>>>> $ sudo apt-get install bob
>>>> This will install bob 1.2.1 that you already have somewhere.
>>>>
>>>> To circumvent the problem you were initially getting when using
>> bob
>>>> 1.2.1 ("Error: Couldn't find a distribution for 'bob==1.2.0'."
>> after
>>>> running the step "bin/buildout"), you could temporarily update
>> your
>>>> setup.py file from the facereclib package as follows:
>>>> Update the line 136:
>>>> "bob == 1.2.0",
>>>> by
>>>> "bob >= 1.2.0",
>>>>
>>>> Then the error message will disappear BUT some of the tests (ISV,
>>
>>>> JFA) from the test suite will fail because of a change in the
>> random
>>>> number generator (cf. issue #116,

>>>> https://github.com/idiap/bob/issues/116 [4] [9]). However, this

>>>>>> https://pypi.python.org/pypi/facereclib/ [5] [5]

>>>>>>
>>>>>> 1. To check if bob is installed properly:
>>>>>> Could you try to open a python interpreter, to run the
>>>>>> following two commands:
>>>>>> >>> import bob
>>>>>> >>> print bob.version
>>>>>> And to tell us the output.
>>>>>>
>>>>>> 2. To test the algorithms provided by the Facereclib, just
>>>>>> follow the instructions here:

>>>>>> https://pypi.python.org/pypi/facereclib [6] [6]

>>>>>>
>>>>>> I'm wondering if the error message you're getting is just
>>>>>> because you are using the facereclib with bob 1.2.1 installed.
>>>>>> The facereclib is currently designed to work against bob 1.2.0,
>>
>>>>>> whereas we just released bob 1.2.1.
>>>>>> A version of the facereclib compatible with bob 1.2.1 will be
>>>>>> available very soon.
>>>>>>
>>>>>> I guess that we could help you further once you will provide
>>>>>> the information above.
>>>>>>
>>>>>> Cheers,
>>>>>> Laurent
>>>>>>
>>>>>> On 17/10/13 10:59, Mathew iprocessor wrote:
>>>>>>
>>>>>>> Hi I have downloaded the Facereclib and followed the steps
>>>>>>> from this link https://pypi.python.org/pypi/facereclib/1.0.0a1

>> [7]

>>>>>>> [1].
>>>>>>>
>>>>>>> But I am getting error as "Error: Couldn't find a distribution
>>
>>>>>>> for 'bob==1.2.0'." after running the step "bin/buildout".
>>>>>>>
>>>>>>> But I have installed BOB also. Still I am facing problem.
>>>>>>>
>>>>>>> 1) How to check whether bob is installed or not? I can see the
>>
>>>>>>> ".so" files in /usr/local/lib
>>>>>>> 2) How to tests the algorithms installed in Facereclib.
>>>>>>>
>>>>>>> I am using ubuntu machine,Detailed steps would be great. --
>>>>>>> -- 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 [8] [2] or
>> directly
>>>>>>> the project website at http://idiap.github.com/bob/ [9] [3]

>>>>>>> ---
>>>>>>> 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.
>>>>>>> For more options, visit

>>>>>>> https://groups.google.com/groups/opt_out [10] [4].
>>>
>>>
>>> Links:
>>> ------
>>> [1] https://pypi.python.org/pypi/facereclib/1.0.0a1 [11]
>>> [2] https://groups.google.com/d/forum/bob-devel [12]
>>> [3] http://idiap.github.com/bob/ [13]
>>> [4] https://groups.google.com/groups/opt_out [14]
>>> [5] https://pypi.python.org/pypi/facereclib/1.0.0a1 [15]
>>> [6] https://pypi.python.org/pypi/facereclib [16]
>>> [7] http://goo.gl/EJJC3 [17]
>>> [8] https://github.com/idiap/bob/wiki/Packages [18]
>>> [9] https://github.com/idiap/bob/issues/116 [19]
>>> [10] https://github.com/bioidiap/xfacereclib.paper.BeFIT2012 [20]

>
> --
> -- 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 [21] or
> directly the project website at http://idiap.github.com/bob/ [22]

> ---
> 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.
> For more options, visit https://groups.google.com/groups/opt_out

> [23].
>
>
> Links:
> ------
> [1] https://github.com/bioidiap/xfacereclib.paper.BeFIT2012
> [2] http://goo.gl/EJJC3
> [3] https://github.com/idiap/bob/wiki/Packages
> [4] https://github.com/idiap/bob/issues/116
> [5] https://pypi.python.org/pypi/facereclib/
> [6] https://pypi.python.org/pypi/facereclib
> [7] https://pypi.python.org/pypi/facereclib/1.0.0a1
> [8] https://groups.google.com/d/forum/bob-devel
> [9] http://idiap.github.com/bob/
> [10] https://groups.google.com/groups/opt_out
> [11] https://pypi.python.org/pypi/facereclib/1.0.0a1
> [12] https://groups.google.com/d/forum/bob-devel
> [13] http://idiap.github.com/bob/
> [14] https://groups.google.com/groups/opt_out
> [15] https://pypi.python.org/pypi/facereclib/1.0.0a1
> [16] https://pypi.python.org/pypi/facereclib
> [17] http://goo.gl/EJJC3
> [18] https://github.com/idiap/bob/wiki/Packages
> [19] https://github.com/idiap/bob/issues/116
> [20] https://github.com/bioidiap/xfacereclib.paper.BeFIT2012
> [21] https://groups.google.com/d/forum/bob-devel
> [22] http://idiap.github.com/bob/
> [23] https://groups.google.com/groups/opt_out

[Mathew iprocessor]

sure.. Thanks

[Mathew iprocessor]

I am little confused. Could you please explain me what is the difference between the below two packages? Sorry for too many questions.

https://pypi.python.org/packages/source/x/xfacereclib.extension.CSU/xfacereclib.extension.CSU-1.0.0.zip and 

https://pypi.python.org/packages/source/x/xfacereclib.paper.BeFIT2012/xfacereclib.paper.BeFT2012-1.0.0.zip 

What suits my requirement? ( As said earlier my requirement is to test the CSU implementations ).

[Laurent El Shafey]

- xfacereclib.extension.CSU is a satellite package to the FaceRecLib that provides wrapper classes for the CSU face recognition resources. Basically, this allows to use the algorithms implemented/provided by CSU as any other toolchain provided by the FaceRecLib, on any database supported by the FaceRecLib.

- xfacereclib.paper.BeFIT2012 is a satellite package to the FaceRecLib that allows to run the experiments published in the following paper:
http://publications.idiap.ch/index.php/publications/show/2431
It has a dependence on xfacereclib.extension.CSU, since it uses the two CSU algorithms in the experiments.

Wrt. to your requirements, I think that you only need xfacereclib.extension.CSU (which depends on the facereclib), as long as you don't want to reproduce the experiments of the paper mentioned above.

Cheers,
Laurent

[Mathew iprocessor]

Thanks for your reply.

xfacereclib.extension.CSU I have downloaded this package - And in this package there is a "setup.py" file, where there is no "bob==1.2.0".

But in "xfacereclib.paper.BeFIT2012" package I could find "bob==1.2.0" in setup.py as you have mentioned in the previous conversation.

Please help. 

[Laurent El Shafey]

True, but then you have to struggle with the facereclib eggs as mentioned earlier. To do that, you have to edit:
eggs/facereclib-1.1.3-py2.7.egg/EGG-INFO/requires.txt and to update "bob == 1.2.0" by "bob >= 1.2.0"
Then, I guess it should be fine.
Please report the error message if any.

But definitely, it would be so much easier if you install bob 1.2.0 from the deb files given earlier, and wait for the packages to be updated to be compatible with bob 1.2.1.

Cheers,
Laurent

[Mathew iprocessor]

Sure, I will try and let you know. Thanks for your patience and support.

[Mathew iprocessor]

I have compiled the package successfully with your instructions(with bob1.2.1). 

But I dont know how to test the CSU algorithms. As said earlier I need to test the face-recognition system with CSU algorithms.

Can you provide me steps to test the CSU algorithm? 

Which is the final binary file which is implemented with CSU algorithm? I could see lot of files in the bin folder. But I couldnt analyse which one to test. Kindly advice.

--
-- 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 a topic in the Google Groups "bob-devel" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/bob-devel/tf0BvlmUig0/unsubscribe.
To unsubscribe from this group and all of its topics, send an email to bob-devel+...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Laurent El Shafey]

You can use the CSU algorithms as any algorithms of the facereclib. The name of the two algorithms are 'lrpca' and 'lda_ir'.
Please consider looking at the documentation of the facereclib for the details.

If you want to get started, just use the bin/baselines.py script, e.g.:
bin/baselines.py -a lrpca

Cheers,
Laurent

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

[Manuel Günther]

Sorry for joining the discussion late, I am currently traveling...

One important fact to note is that we do not provide the original source code for the LRPCA and CohortLDA (we use its old name LDA-IR), but we only provide wrapper classes for these packages to be used with the FaceRecLib. For more details on downloading and installing the CSU resources, please read:

http://pypi.python.org/pypi/facereclib

or

http://pypi.python.org/pypi/xfacereclib.extension.CSU

After the installation, you need to set up your database configuration to be able to run experiments on the database using the 'bin/baselines.py' command line utility that Laurent mentioned. For more details, please read the documentation of the FaceRecLib (see http://pypi.python.org/pypi/facereclib for instruction how to generate the documentation).

In case you have more questions or comments, you are highly welcome to post here. This helps us, e.g., in improving the documentation.

Cheers

Manuel

[Mathew iprocessor]

I followed these steps here https://pypi.python.org/pypi/xfacereclib.extension.CSU and when I ran the step  "bin/nosetests", I got the below error. It shows as "failures=2". What would be the reason?

test01_ldair_preprocessing (xfacereclib.extension.CSU.tests.PythonFaceEvaluationTest) ... ok

test02_ldair_features (xfacereclib.extension.CSU.tests.PythonFaceEvaluationTest) ... ok

test03_ldair_tool (xfacereclib.extension.CSU.tests.PythonFaceEvaluationTest) ... ok

test04_lrpca_preprocessing (xfacereclib.extension.CSU.tests.PythonFaceEvaluationTest) ... FAIL

test05_lrpca_features (xfacereclib.extension.CSU.tests.PythonFaceEvaluationTest) ... FAIL

test06_lrpca_tool (xfacereclib.extension.CSU.tests.PythonFaceEvaluationTest) ... ok

======================================================================

FAIL: test04_lrpca_preprocessing (xfacereclib.extension.CSU.tests.PythonFaceEvaluationTest)

----------------------------------------------------------------------

Traceback (most recent call last):

  File "/home/tsk/xfacereclib.extension.CSU-1.0.0/xfacereclib/extension/CSU/tests.py", line 166, in test04_lrpca_preprocessing

    self.execute_preprocessor(preprocessor, image, annotations, 'lrpca.hdf5')

  File "/home/tsk/xfacereclib.extension.CSU-1.0.0/xfacereclib/extension/CSU/tests.py", line 64, in execute_preprocessor

    self.assertTrue((numpy.abs(preprocessor.read_data(self.input('preprocessing', reference)) - preprocessed) < 1e-5).all())

AssertionError: False is not true

======================================================================

FAIL: test05_lrpca_features (xfacereclib.extension.CSU.tests.PythonFaceEvaluationTest)

----------------------------------------------------------------------

Traceback (most recent call last):

  File "/home/tsk/xfacereclib.extension.CSU-1.0.0/xfacereclib/extension/CSU/tests.py", line 201, in test05_lrpca_features

    feature = self.execute_extractor(extractor, image, 'lrpca.hdf5')

  File "/home/tsk/xfacereclib.extension.CSU-1.0.0/xfacereclib/extension/CSU/tests.py", line 73, in execute_extractor

    self.assertTrue((numpy.abs(bob.io.load(self.input('features', reference)) - feature) < 1e-5).all())

AssertionError: False is not true

----------------------------------------------------------------------

Ran 6 tests in 16.497s

FAILED (failures=2)

Meanwhile, my intention is to do a face-verification system and below is the description of my application.

1.I have "Un" no of users.

2.For each users (U1,U2,....Un),I will get some 5 images for enrolment. I will store the images in a server environment under each 

user's folder(Training/Testing).

3.When user attempts to get access to the system, I will capture one or more images.I will store them as well in a server along with 

enrollment images. 

Can I do this face verification-system as explained above with the scripts present in the facereclib package? If so, Can 

you please help me out by providing some tips ans steps.

Thanks,

Mathew

[Laurent El Shafey]

Hello,

Wrt. the tests failing, I have no explanation why. Maybe Manuel will be able to help you for that, since he is the author of this package.

If you want to run face verification experiments on your own data, you can use the xbob.db.verification.filelist satellite package (http://pypi.python.org/pypi/xbob.db.verification.filelist).
To do that, in your buidout.cfg file, you have to add a line in the list of eggs with
      xbob.db.verification.filelist
and to rerun the bin/buildout script (this should download this additional package). Next you just have to define your own lists (training, enrollment and probe) for a protocol, and to create a database configuration file for this protocol (look at the facereclib/configurations/databases/ repository for examples).

If you don't want to 'evaluate' a face verification system, but rather to 'deploy' a face verification system, I guess that you will have to update a bit the scripts from the facereclib. However, it might still be possible to do it, by using the bin/faceverify.py script and the --skip-* optional arguments.In addition, if you just want to deploy a face verification system based on the CSU algorithms, I'm not sure that using the xfacereclib.extension.CSU package would be of any help. Just tweak the CSU source code directly.

Cheers,
Laurent

[Manuel Günther]

Dear Mathew,

actually I doubt that the FaceRecLib is the proper tool for your purpose. The FaceRecLib is designed to run a fair comparison of face recognition algorithms on some default image databases. It is *not* designed to be used in an online system like yours. Hence, scripts for your purpose do not exist in the FaceRecLib.

I would rather suggest that you build your system around the original CSU framework, using their code directly, and not through the FaceRecLib. This is probably much easier for you.

Best regards
Manuel

[Mathew iprocessor]

Thanks for your reply and support. It really helps me a lot. Regarding face-recognition algorithms , I have gone through different algorithms, and heard that OpenBr, Inter-Session Variability Modeling, and CSU are few among the best algorithms which are opensource too. From your perspective and expertise, can you suggest any other good or latest face-recognition algorithms that outperform any of these three?

Awaiting your reply..

Thanks,

Mathew

--

[Manuel Günther]

Dear Mathew,

unfortunately, there is no "easy" answer for your question, it highly depends on the type of images that you have. For more information, you might want to read a post on the open-br group that I wrote some days ago:

http://groups.google.com/forum/#!topic/openbr-dev/HWMkL4rL3Lg

Best regards
Manuel

[Mathew iprocessor]

My typical project environment:

As mentioned earlier, my intention is to do a face-verification system which receives "nq" query images at the time of registration or enrolment and stores them in a server under an unique folder names,for each user. And at the time of login I will collect "nt" images as test images for authentication and store them as well.

Decision Making:

Then, I will calculate the matching score and make a decision on deciding the "saturation value". I hope this is the correct and the best way. Am I right?

Pasted from your post with my in-line questions. To save your time, kindly answer in-line.

unfortunately, there is no simple answer to your question. My recommendation depends highly on the actual application which it is used for:

- When time and memory is not important, I would *definitely* suggest the ISV algorithm since it was able to outperform all the other algorithms I have tested so far. But, as indicated this algorithm needs are very long training time and still a quite long test time. Also, the DCT features that we usually use are a bit prone to for illumination.

Do you mean the processing time? Could you elaborate ? 

Is this algorithm available for public to use freely?

Keeping high accuracy in mind and considering my project environment, can I use this algorithm?

Could you define this algorithm's accuracy in terms of % comparing other algorithms?

I will be handling all the operations in servers by capturing images through mobile or web browsers.

- When illumination is uncontrolled, I would suggest an algorithm based on Gabor filter responses, as Gabor graph comparisons or local Gabor binary pattern histogram sequences. Both algorithms have moderate speed and memory requirements

.
- When you can control the illumination and people cooperate, there are many algorithms showing good performance. When speed is required, I would use something simple like LDA.

In what are all the ways, I can ask the users to co-operate with the system to get the maximum accuracy, apart from instructing them to provide 2d frontal face images in good lighting conditions? 

- With low resolution or bad quality data, the CohortLDA algorithm, which are implemented in the CSU Face Recognition Resources and which are wrapped in the FaceRecLib, that uses color information seems to give the best results.

   I am reading about this algorithm.

I can also combine 1+ algorithms as i handle everything at server?
Any general advice for me and my project to improve the accuracy?

--

[Mathew iprocessor]

Gentle reminder!

--
-- 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 a topic in the Google Groups "bob-devel" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/bob-devel/tf0BvlmUig0/unsubscribe.

To unsubscribe from this group and all of its topics, send an email to bob-devel+unsubscribe@googlegroups.com.

[Sebastien Marcel]

Dear Mathew

thank you for your interest

All the algorithms implemented in Bob (including ISV) and executed from face rec lib are fee of use under the terms of the GPL license. Please check it on the website.

A lot of your questions are mostly general questions on face recognition (complexity, robustness) and for this there is no single simple answer.

I can only recommend you to read our papers on face recognition (or from others) where you will find the answers to all your questions:

Our publication can be found here: http://www.idiap.ch/~marcel/professional/Publications.html#topics:Face%20Recognition

A tutorial on face recognition can be found here: http://www.idiap.ch/~marcel/professional/UNICA_2010.html

Within this google group we try to provide the best support we can with the limited resources we have to successfully install and use Bob + its satellite packages.

Other questions on the use/exploitation of the technology (face recognition for example) that require more in depth analysis and elaboration are more related to consulting or to technology

licensing. This should be discussed with me by email ( mar...@idiap.ch ).

best

Sebastien

--
-- 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 a topic in the Google Groups "bob-devel" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/bob-devel/tf0BvlmUig0/unsubscribe.

To unsubscribe from this group and all of its topics, send an email to bob-devel+...@googlegroups.com.