[MINC-users] classify: how to use

[Alex Zijdenbos]

Nice, Jason! Next time I see Vasco I'll let him know ;-) I thought that
crashing bit was fixed somewhere along the way, but perhaps not - haven't
tried in a while.

To answer Pedro's initial question though - the most common way to invoke
good old classify directly would look something like this:

   classify [options] -<classifier> -tagfile <tags.tag> <in.mnc> [<in.mnc>
...] <classified.mnc>

where you should only need to provide a tag file with training samples for
the supervised classifiers. classify_clean, as Andrew suggested, runs two
of these calls (minimum distance followed by ANN) so its logfile will also
help you.

-- A




On Wed, May 4, 2016 at 7:14 AM, Jason Lerch <jason at mouseimaging.ca> wrote:

> Hi all,
>
> for those interested, we created a sort of replacement for classify:
>
> https://github.com/Mouse-Imaging-Centre/classify-sklearn <
> https://github.com/Mouse-Imaging-Centre/classify-sklearn>
>
> It’s based on python’s scikit-learn, and currently comes with a mix of
> classifiers (random forests, bayes, etc.). An example or two of how to use
> it is on the wiki part of that github page (though obviously more
> documentation is needed).
>
> Biggest advantage over classify - doesn’t crash when outputting posterior
> probabilities with multispectral input - and doesn’t require the creation
> of a frankenbrain when training a classifier based on multiple subjects.
> Otherwise should be quite similar to the venerable old classify.
>
> Doesn’t have a variant of classify_clean yet though.
>
> Jason
>
> > On May 3, 2016, at 7:21 PM, Andrew Janke <a.janke at gmail.com> wrote:
> >
> > Hi Pedro,
> >
> > This is my approach:
> >
> >   classify_clean \
> >         -tagdir <tags> \
> >         -tag_transform <image-to-tags.xfm> \
> >         input.mnc output.mnc
> >
> > <tags> is a directory with the classify tags in it eg:
> >
> > ls tags/
> > ntags_1000_bg.tag  ntags_1000_prob_90_nobg.tag
> >
> >
> > <image-to-tags.xfm> is a tranform from your native data space
> > (input.mnc) to the space of the tags. If you are using the ones in the
> > standard package this will be a transform from native space to the
> > icbm152 model.
> >
> >
> > a
> >
> >
> >
> > On 3 May 2016 at 23:36, Pedro <ptcougopinto at gmail.com> wrote:
> >>
> >> Hi, everyone —
> >>
> >> What should be the minimum command line for classify? What are the
> inputs and output? I could´t figure out from -help. (Just found out that
> I´m not happy with registration based segmentation of my data).
> >>
> >> Thanks!
> >> Pedro
> >> _______________________________________________
> >> MINC-users at bic.mni.mcgill.ca
> >> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
> > _______________________________________________
> > MINC-users at bic.mni.mcgill.ca
> > http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>
> _______________________________________________
> MINC-users at bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>