Discussion:
Advanced Weka Segmentation
Christian Tischer
2012-05-10 13:07:23 UTC
Permalink
Hello,

i have a question about the (fantastic!) Advanced Weka Segmentation:

Are the "Settings" (i.e. which features to compute) stored together with
the classifier file?

i suppose this is the case, because i guess each classifier is strongly
linked to the features, however, after "Load classifier" the feature
entries in "Settings" are not changed/updated. could it be that they are
only changed internally but not visibly updated?

in fact the class names are updated (so this seems to behave as expected..)

could someone maybe explain me how this works?

thanks!!!

Tischi
--
Please avoid top-posting, and please make sure to reply-to-all!

Mailing list web interface: http://groups.google.com/group/fiji-users
Ignacio Arganda-Carreras
2012-05-10 13:33:27 UTC
Permalink
Hello Christian,

Yes, the feature names are stored in the classifier. So every time you load
a new classifier from file the plugin will read the feature names, which
also include information about the radius of the filters to use.

Let me know if need any help!

ignacio

On Thu, May 10, 2012 at 9:07 AM, Christian Tischer <
Post by Christian Tischer
Hello,
Are the "Settings" (i.e. which features to compute) stored together with
the classifier file?
i suppose this is the case, because i guess each classifier is strongly
linked to the features, however, after "Load classifier" the feature
entries in "Settings" are not changed/updated. could it be that they are
only changed internally but not visibly updated?
in fact the class names are updated (so this seems to behave as expected..)
could someone maybe explain me how this works?
thanks!!!
Tischi
--
Please avoid top-posting, and please make sure to reply-to-all!
Mailing list web interface: http://groups.google.com/group/fiji-users
--
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://biocomp.cnb.csic.es/~iarganda/index_EN.html
--
Please avoid top-posting, and please make sure to reply-to-all!

Mailing list web interface: http://groups.google.com/group/fiji-users
Christian Tischer
2012-05-10 13:41:32 UTC
Permalink
Hi Ignacio,

-so would you say it is a "bug" that the Settings are not updated?!

in fact i have a another question:

- is it possible to only "train" the classifier without(!) automatically
"applying" it afterward??
(the problem is that we select just a few examples out of a stack of
hundreds of images, and we actually do not want to classify all these
images (in fact the computer crashed if we do so))

if i dare may make a suggestion: what about making it a user-choice in the
Settings, where in the function:

void runStopTraining(final String command)

the following part of the code is not executed if the User does not want it:

....
wekaSegmentation.applyClassifier(false);
classifiedImage = wekaSegmentation.getClassifiedImage();
if(showColorOverlay)
win.toggleOverlay();
win.toggleOverlay();
....


Cheers, Tischi

PS: i sent the latter questions already to some private e-mail account of
yours, but maybe that was an old one ;-)
--
Please avoid top-posting, and please make sure to reply-to-all!

Mailing list web interface: http://groups.google.com/group/fiji-users
Ignacio Arganda-Carreras
2012-05-10 13:57:30 UTC
Permalink
Hello Christian,
Post by Christian Tischer
-so would you say it is a "bug" that the Settings are not updated?!
Yes, that's a bug! Thanks for reporting, i'll work on it :)
Post by Christian Tischer
- is it possible to only "train" the classifier without(!) automatically
"applying" it afterward??
(the problem is that we select just a few examples out of a stack of
hundreds of images, and we actually do not want to classify all these
images (in fact the computer crashed if we do so))
Right now, there's no way of doing that using the GUI, only through the
library methods in WekaSegmentation.java. Why would you like such a feature
in the plugin? It goes a bit against the whole interactive learning idea
behind Advanced Weka Segmentation.

Just let me know what and why you need it and we can try to put the feature
there ;)

ignacio
--
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://biocomp.cnb.csic.es/~iarganda/index_EN.html
--
Please avoid top-posting, and please make sure to reply-to-all!

Mailing list web interface: http://groups.google.com/group/fiji-users
Loading...