Login  Register

Problem with Weka Trainable Classifier

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
5 messages Options Options
Embed post
Permalink
Reply | Threaded
Open this post in threaded view
| More
Print post
Permalink

Problem with Weka Trainable Classifier

Sidnei Paciornik
Dear All,

I have a strange behavior with the Weka Trainable Classifier.

I have:
- Trained a classifier - just fine
- Created a result - the segmented image is correctly displayed
- Saved the classifier (probably unnecessary)
- Loaded the classifier (probably unnecessary as there was an active
classifier)
- When I choose Apply Classifier the system seems to do the job:
    - It opens the dialog for me to select an image from a folder,
    - asks if I want to calculate probabilities instead of segmentation (I
answer no)
    - the log window displays some information,

but the segmented image never shows up.

I am running under Windows 7 Home 64-bits, ImageJ 1.47v and have just
updated FIJI. I got the same behavior under Windows XP and Linux.

What am I doing wrong?

Thank you!

Prof. Sidnei Paciornik
Grupo de Análise de Imagens e Microscopia Digital
DEMa <http://www.dema.puc-rio.br/> - Departamento de Engenharia de Materiais
PUC-Rio <http://www.puc-rio.br/>
Rua Marquês de São Vicente 225
Prédio Leme, Sala 501L
Gávea - Rio de Janeiro - RJ
22451-900 - Brasil
tel: (55)(21)3527-1243

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Reply | Threaded
Open this post in threaded view
| More
Print post
Permalink

Re: Problem with Weka Trainable Classifier

Ignacio Arganda-Carreras
Hello Sidnei,

Thanks for your interest in the plugin. Let me answer you between your
lines:

I have:
> - Trained a classifier - just fine
> - Created a result - the segmented image is correctly displayed
> - Saved the classifier (probably unnecessary)
>

Well, it is always a good idea if you liked your results. This way you can
reuse the classifier later on other images.


> - Loaded the classifier (probably unnecessary as there was an active
> classifier)
>

No, that part was not necessary. As you said, the classifier is already
active.


> - When I choose Apply Classifier the system seems to do the job:
>     - It opens the dialog for me to select an image from a folder,
>     - asks if I want to calculate probabilities instead of segmentation (I
> answer no)
>     - the log window displays some information,
>

Can you send me the information that is displayed in the Low window,
please? That should help me understand how far the program went.


>
> but the segmented image never shows up.
>
>
And I assume you didn't get any error message, right? That's sounds like a
bug... or a memory problem.



> I am running under Windows 7 Home 64-bits, ImageJ 1.47v and have just
> updated FIJI. I got the same behavior under Windows XP and Linux.
>
> What am I doing wrong?
>

How much RAM memory do you have and how large your images are? That can be
the key.

If you want, you can send me your images so i can test on them myself.

Thanks for reporting!

ignacio



>
> Thank you!
>
> Prof. Sidnei Paciornik
> Grupo de Análise de Imagens e Microscopia Digital
> DEMa <http://www.dema.puc-rio.br/> - Departamento de Engenharia de
> Materiais
> PUC-Rio <http://www.puc-rio.br/>
> Rua Marquês de São Vicente 225
> Prédio Leme, Sala 501L
> Gávea - Rio de Janeiro - RJ
> 22451-900 - Brasil
> tel: (55)(21)3527-1243
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>



--
Ignacio Arganda-Carreras, Ph.D.
Seung's lab, 46-5065
Department of Brain and Cognitive Sciences
Massachusetts Institute of Technology
43 Vassar St.
Cambridge, MA 02139
USA

Phone: (001) 617-324-3747
Website: http://bioweb.cnb.csic.es/~iarganda/index_EN.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Reply | Threaded
Open this post in threaded view
| More
Print post
Permalink

Re: Problem with Weka Trainable Classifier

Sidnei Paciornik
Hi Ignacio,

Thank you for your detailed feedback.

It is really strange. I just tried to reproduce the error in the same
Win7-64 machine and it did not occur. It runs fine and shows the resulting
segmented image, besides opening the input image.

At first I did thing you were right and that it was a memory problem. I
tried it with the small blobs.gif image and it worked. I then moved on to
the image that had given me trouble, and it worked too! It is really weird
because the same bad behavior had occurred in different computers and OS's.

The only environment changes in the Win7-64 machine between the two
situations (bad and then good) were;
1 - The bad behaviors occurred with the computer connected to a network
that requires a proxy. The good behavior occurred with the computer
connected via wifi without a proxy. Does the system depend on internet
connection at all?? It does not seem to make sense...

2 - In Win7-64 bits Sony Vaio, which gave the bad result before, there was
a major Windows update in between the trials ...I am not sure what got
installed. Could it affect the behavior of the plugin? Again, sounds
unlikely, except that the computer was restarted.

For that matter, this Sony Vaio has 8 Gb of RAM.

So I don't know what is going on...I will make some more tests and let you
know if I find out something.

BTW, the method and the plugin are REALLY impressive! I've got results with
it that seemed almost impossible with more traditional segmentation
approaches.

Thank you again.

Prof. Sidnei Paciornik
Grupo de Análise de Imagens e Microscopia Digital
DEMa <http://www.dema.puc-rio.br/> - Departamento de Engenharia de Materiais
PUC-Rio <http://www.puc-rio.br/>
Rua Marquês de São Vicente 225
Prédio Leme, Sala 501L
Gávea - Rio de Janeiro - RJ
22451-900 - Brasil
tel: (55)(21)3527-1243





On Thu, Jul 11, 2013 at 4:48 PM, Ignacio Arganda-Carreras <
[hidden email]> wrote:

> Hello Sidnei,
>
> Thanks for your interest in the plugin. Let me answer you between your
> lines:
>
> I have:
> > - Trained a classifier - just fine
> > - Created a result - the segmented image is correctly displayed
> > - Saved the classifier (probably unnecessary)
> >
>
> Well, it is always a good idea if you liked your results. This way you can
> reuse the classifier later on other images.
>
>
> > - Loaded the classifier (probably unnecessary as there was an active
> > classifier)
> >
>
> No, that part was not necessary. As you said, the classifier is already
> active.
>
>
> > - When I choose Apply Classifier the system seems to do the job:
> >     - It opens the dialog for me to select an image from a folder,
> >     - asks if I want to calculate probabilities instead of segmentation
> (I
> > answer no)
> >     - the log window displays some information,
> >
>
> Can you send me the information that is displayed in the Low window,
> please? That should help me understand how far the program went.
>
>
> >
> > but the segmented image never shows up.
> >
> >
> And I assume you didn't get any error message, right? That's sounds like a
> bug... or a memory problem.
>
>
>
> > I am running under Windows 7 Home 64-bits, ImageJ 1.47v and have just
> > updated FIJI. I got the same behavior under Windows XP and Linux.
> >
> > What am I doing wrong?
> >
>
> How much RAM memory do you have and how large your images are? That can be
> the key.
>
> If you want, you can send me your images so i can test on them myself.
>
> Thanks for reporting!
>
> ignacio
>
>
>
> >
> > Thank you!
> >
> > Prof. Sidnei Paciornik
> > Grupo de Análise de Imagens e Microscopia Digital
> > DEMa <http://www.dema.puc-rio.br/> - Departamento de Engenharia de
> > Materiais
> > PUC-Rio <http://www.puc-rio.br/>
> > Rua Marquês de São Vicente 225
> > Prédio Leme, Sala 501L
> > Gávea - Rio de Janeiro - RJ
> > 22451-900 - Brasil
> > tel: (55)(21)3527-1243
> >
> > --
> > ImageJ mailing list: http://imagej.nih.gov/ij/list.html
> >
>
>
>
> --
> Ignacio Arganda-Carreras, Ph.D.
> Seung's lab, 46-5065
> Department of Brain and Cognitive Sciences
> Massachusetts Institute of Technology
> 43 Vassar St.
> Cambridge, MA 02139
> USA
>
> Phone: (001) 617-324-3747
> Website: http://bioweb.cnb.csic.es/~iarganda/index_EN.html
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Reply | Threaded
Open this post in threaded view
| More
Print post
Permalink

Re: Problem with Weka Trainable Classifier

Ignacio Arganda-Carreras
Hello again, Sidnei!

It is really strange. I just tried to reproduce the error in the same
> Win7-64 machine and it did not occur. It runs fine and shows the resulting
> segmented image, besides opening the input image.
>
>
Weird!


> At first I did thing you were right and that it was a memory problem. I
> tried it with the small blobs.gif image and it worked. I then moved on to
> the image that had given me trouble, and it worked too! It is really weird
> because the same bad behavior had occurred in different computers and OS's.


And now the error does not happen in ANY of them?


>

The only environment changes in the Win7-64 machine between the two
> situations (bad and then good) were;
> 1 - The bad behaviors occurred with the computer connected to a network
> that requires a proxy. The good behavior occurred with the computer
> connected via wifi without a proxy. Does the system depend on internet
> connection at all?? It does not seem to make sense...
>

No. Applying the classifier that you trained (or loaded) does not need any
connection. It's very strange, but you're not the only one experiencing
this behavior so we're probably missing something.


>
> 2 - In Win7-64 bits Sony Vaio, which gave the bad result before, there was
> a major Windows update in between the trials ...I am not sure what got
> installed. Could it affect the behavior of the plugin? Again, sounds
> unlikely, except that the computer was restarted.
>

Unless it had to do with java, I doubt the update had an effect in the
plugin. Fiji uses his own java jre and that's what you're using, right?

The next time you have a problem like this, can you please try something?
If the loaded classifier get's stuck when clicking on "Apply classifier",
please close Fiji and open it again and this time open the plugin with the
image you wanted to have segmented. Then load the classifier and click on
"Create result" or "Get probability" and let me know if that works.

The only different would be in the way the threads are used (one thread per
image if you click on "Apply classifier" and all threads for each image in
a sequence if you use the other options). That's the only origin of the
problem I can think of.


>
> For that matter, this Sony Vaio has 8 Gb of RAM.
>
> So I don't know what is going on...I will make some more tests and let you
> know if I find out something.
>

Great, thanks for reporting, that really helps me improving the performance
and stability of the plugin.


>
> BTW, the method and the plugin are REALLY impressive! I've got results with
> it that seemed almost impossible with more traditional segmentation
> approaches.
>
>
I'm glad to hear that!


> Thank you again.
>
>
My pleasure!

ignacio


>  Prof. Sidnei Paciornik
> Grupo de Análise de Imagens e Microscopia Digital
> DEMa <http://www.dema.puc-rio.br/> - Departamento de Engenharia de
> Materiais
> PUC-Rio <http://www.puc-rio.br/>
> Rua Marquês de São Vicente 225
> Prédio Leme, Sala 501L
> Gávea - Rio de Janeiro - RJ
> 22451-900 - Brasil
> tel: (55)(21)3527-1243
>
>
>
>
>
> On Thu, Jul 11, 2013 at 4:48 PM, Ignacio Arganda-Carreras <
> [hidden email]> wrote:
>
> > Hello Sidnei,
> >
> > Thanks for your interest in the plugin. Let me answer you between your
> > lines:
> >
> > I have:
> > > - Trained a classifier - just fine
> > > - Created a result - the segmented image is correctly displayed
> > > - Saved the classifier (probably unnecessary)
> > >
> >
> > Well, it is always a good idea if you liked your results. This way you
> can
> > reuse the classifier later on other images.
> >
> >
> > > - Loaded the classifier (probably unnecessary as there was an active
> > > classifier)
> > >
> >
> > No, that part was not necessary. As you said, the classifier is already
> > active.
> >
> >
> > > - When I choose Apply Classifier the system seems to do the job:
> > >     - It opens the dialog for me to select an image from a folder,
> > >     - asks if I want to calculate probabilities instead of segmentation
> > (I
> > > answer no)
> > >     - the log window displays some information,
> > >
> >
> > Can you send me the information that is displayed in the Low window,
> > please? That should help me understand how far the program went.
> >
> >
> > >
> > > but the segmented image never shows up.
> > >
> > >
> > And I assume you didn't get any error message, right? That's sounds like
> a
> > bug... or a memory problem.
> >
> >
> >
> > > I am running under Windows 7 Home 64-bits, ImageJ 1.47v and have just
> > > updated FIJI. I got the same behavior under Windows XP and Linux.
> > >
> > > What am I doing wrong?
> > >
> >
> > How much RAM memory do you have and how large your images are? That can
> be
> > the key.
> >
> > If you want, you can send me your images so i can test on them myself.
> >
> > Thanks for reporting!
> >
> > ignacio
> >
> >
> >
> > >
> > > Thank you!
> > >
> > > Prof. Sidnei Paciornik
> > > Grupo de Análise de Imagens e Microscopia Digital
> > > DEMa <http://www.dema.puc-rio.br/> - Departamento de Engenharia de
> > > Materiais
> > > PUC-Rio <http://www.puc-rio.br/>
> > > Rua Marquês de São Vicente 225
> > > Prédio Leme, Sala 501L
> > > Gávea - Rio de Janeiro - RJ
> > > 22451-900 - Brasil
> > > tel: (55)(21)3527-1243
> > >
> > > --
> > > ImageJ mailing list: http://imagej.nih.gov/ij/list.html
> > >
> >
> >
> >
> > --
> > Ignacio Arganda-Carreras, Ph.D.
> > Seung's lab, 46-5065
> > Department of Brain and Cognitive Sciences
> > Massachusetts Institute of Technology
> > 43 Vassar St.
> > Cambridge, MA 02139
> > USA
> >
> > Phone: (001) 617-324-3747
> > Website: http://bioweb.cnb.csic.es/~iarganda/index_EN.html
> >
> > --
> > ImageJ mailing list: http://imagej.nih.gov/ij/list.html
> >
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>



--
Ignacio Arganda-Carreras, Ph.D.
Seung's lab, 46-5065
Department of Brain and Cognitive Sciences
Massachusetts Institute of Technology
43 Vassar St.
Cambridge, MA 02139
USA

Phone: (001) 617-324-3747
Website: http://bioweb.cnb.csic.es/~iarganda/index_EN.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Reply | Threaded
Open this post in threaded view
| More
Print post
Permalink

Re: Problem with Weka Trainable Classifier

Sidnei Paciornik
Hi Ignacio,

I only tried again with the Sony Vaio Win7-Home-64 bits-8Gb RAM. I have not
yet checked with the other computers and OS's. I should be able to do that
on Monday.

I did no changes regarding the use of Java. So it is the default
configuration used by FIJI/ImageJ.

Thank you again and let us keep on testing.

Regards,

Sidnei


On Fri, Jul 12, 2013 at 10:46 AM, Ignacio Arganda-Carreras <
[hidden email]> wrote:

> Hello again, Sidnei!
>
> It is really strange. I just tried to reproduce the error in the same
> > Win7-64 machine and it did not occur. It runs fine and shows the
> resulting
> > segmented image, besides opening the input image.
> >
> >
> Weird!
>
>
> > At first I did thing you were right and that it was a memory problem. I
> > tried it with the small blobs.gif image and it worked. I then moved on to
> > the image that had given me trouble, and it worked too! It is really
> weird
> > because the same bad behavior had occurred in different computers and
> OS's.
>
>
> And now the error does not happen in ANY of them?
>
>
> >
>
> The only environment changes in the Win7-64 machine between the two
> > situations (bad and then good) were;
> > 1 - The bad behaviors occurred with the computer connected to a network
> > that requires a proxy. The good behavior occurred with the computer
> > connected via wifi without a proxy. Does the system depend on internet
> > connection at all?? It does not seem to make sense...
> >
>
> No. Applying the classifier that you trained (or loaded) does not need any
> connection. It's very strange, but you're not the only one experiencing
> this behavior so we're probably missing something.
>
>
> >
> > 2 - In Win7-64 bits Sony Vaio, which gave the bad result before, there
> was
> > a major Windows update in between the trials ...I am not sure what got
> > installed. Could it affect the behavior of the plugin? Again, sounds
> > unlikely, except that the computer was restarted.
> >
>
> Unless it had to do with java, I doubt the update had an effect in the
> plugin. Fiji uses his own java jre and that's what you're using, right?
>
> The next time you have a problem like this, can you please try something?
> If the loaded classifier get's stuck when clicking on "Apply classifier",
> please close Fiji and open it again and this time open the plugin with the
> image you wanted to have segmented. Then load the classifier and click on
> "Create result" or "Get probability" and let me know if that works.
>
> The only different would be in the way the threads are used (one thread per
> image if you click on "Apply classifier" and all threads for each image in
> a sequence if you use the other options). That's the only origin of the
> problem I can think of.
>
>
> >
> > For that matter, this Sony Vaio has 8 Gb of RAM.
> >
> > So I don't know what is going on...I will make some more tests and let
> you
> > know if I find out something.
> >
>
> Great, thanks for reporting, that really helps me improving the performance
> and stability of the plugin.
>
>
> >
> > BTW, the method and the plugin are REALLY impressive! I've got results
> with
> > it that seemed almost impossible with more traditional segmentation
> > approaches.
> >
> >
> I'm glad to hear that!
>
>
> > Thank you again.
> >
> >
> My pleasure!
>
> ignacio
>
>
> >  Prof. Sidnei Paciornik
> > Grupo de Análise de Imagens e Microscopia Digital
> > DEMa <http://www.dema.puc-rio.br/> - Departamento de Engenharia de
> > Materiais
> > PUC-Rio <http://www.puc-rio.br/>
> > Rua Marquês de São Vicente 225
> > Prédio Leme, Sala 501L
> > Gávea - Rio de Janeiro - RJ
> > 22451-900 - Brasil
> > tel: (55)(21)3527-1243
> >
> >
> >
> >
> >
> > On Thu, Jul 11, 2013 at 4:48 PM, Ignacio Arganda-Carreras <
> > [hidden email]> wrote:
> >
> > > Hello Sidnei,
> > >
> > > Thanks for your interest in the plugin. Let me answer you between your
> > > lines:
> > >
> > > I have:
> > > > - Trained a classifier - just fine
> > > > - Created a result - the segmented image is correctly displayed
> > > > - Saved the classifier (probably unnecessary)
> > > >
> > >
> > > Well, it is always a good idea if you liked your results. This way you
> > can
> > > reuse the classifier later on other images.
> > >
> > >
> > > > - Loaded the classifier (probably unnecessary as there was an active
> > > > classifier)
> > > >
> > >
> > > No, that part was not necessary. As you said, the classifier is already
> > > active.
> > >
> > >
> > > > - When I choose Apply Classifier the system seems to do the job:
> > > >     - It opens the dialog for me to select an image from a folder,
> > > >     - asks if I want to calculate probabilities instead of
> segmentation
> > > (I
> > > > answer no)
> > > >     - the log window displays some information,
> > > >
> > >
> > > Can you send me the information that is displayed in the Low window,
> > > please? That should help me understand how far the program went.
> > >
> > >
> > > >
> > > > but the segmented image never shows up.
> > > >
> > > >
> > > And I assume you didn't get any error message, right? That's sounds
> like
> > a
> > > bug... or a memory problem.
> > >
> > >
> > >
> > > > I am running under Windows 7 Home 64-bits, ImageJ 1.47v and have just
> > > > updated FIJI. I got the same behavior under Windows XP and Linux.
> > > >
> > > > What am I doing wrong?
> > > >
> > >
> > > How much RAM memory do you have and how large your images are? That can
> > be
> > > the key.
> > >
> > > If you want, you can send me your images so i can test on them myself.
> > >
> > > Thanks for reporting!
> > >
> > > ignacio
> > >
> > >
> > >
> > > >
> > > > Thank you!
> > > >
> > > > Prof. Sidnei Paciornik
> > > > Grupo de Análise de Imagens e Microscopia Digital
> > > > DEMa <http://www.dema.puc-rio.br/> - Departamento de Engenharia de
> > > > Materiais
> > > > PUC-Rio <http://www.puc-rio.br/>
> > > > Rua Marquês de São Vicente 225
> > > > Prédio Leme, Sala 501L
> > > > Gávea - Rio de Janeiro - RJ
> > > > 22451-900 - Brasil
> > > > tel: (55)(21)3527-1243
> > > >
> > > > --
> > > > ImageJ mailing list: http://imagej.nih.gov/ij/list.html
> > > >
> > >
> > >
> > >
> > > --
> > > Ignacio Arganda-Carreras, Ph.D.
> > > Seung's lab, 46-5065
> > > Department of Brain and Cognitive Sciences
> > > Massachusetts Institute of Technology
> > > 43 Vassar St.
> > > Cambridge, MA 02139
> > > USA
> > >
> > > Phone: (001) 617-324-3747
> > > Website: http://bioweb.cnb.csic.es/~iarganda/index_EN.html
> > >
> > > --
> > > ImageJ mailing list: http://imagej.nih.gov/ij/list.html
> > >
> >
> > --
> > ImageJ mailing list: http://imagej.nih.gov/ij/list.html
> >
>
>
>
> --
> Ignacio Arganda-Carreras, Ph.D.
> Seung's lab, 46-5065
> Department of Brain and Cognitive Sciences
> Massachusetts Institute of Technology
> 43 Vassar St.
> Cambridge, MA 02139
> USA
>
> Phone: (001) 617-324-3747
> Website: http://bioweb.cnb.csic.es/~iarganda/index_EN.html
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html