Re: Cytoseg on cluster0 update

[Rick Giuly]

Hi,

I think it looks good, but I'd need to see the block that you are using
for the corresponding data (the actual images) to see if they line up.
The training seg block and the data block that goes with it should have
the exact same dimensions.

Also,

There's a config file that specifies what type of object the numeric
values in your training data correspond to.

It's cytoseg/testing/sbfsem_settings1.py

If you want 0 to be "not microtubules" and nonzero to be microtubules,
something like this will work (I just edited two lines):

from label_identifier import *

def mapNumbersToComponents(detector):

for object in (detector.contourClassifier,):

object.labelIdentifierDict['microtubules'] =\
LabelIdentifier(min=1, max=255)
object.labelIdentifierDict['other'] =\
LabelIdentifier(min=0, max=0)

#todo: this shouldn't be here
detector.setTarget('mitochondria_new')

-Rick

Masako Terada wrote:
> /ncmirdata4/mterada/for_ying2/Dauertomo2_20k/cytoseg
>
> Dauer2_segmentation.rec
>
> I made a test segmentation in AMIRA. Any suggestions?
>