open-discussion
1 Subscriber 
open-discussion > Corrupted spatial maps
Showing 1-2 of 2 posts
Apr 8, 2011 06:04 PM | Alexander Lebedev
Corrupted spatial maps
Dear COMPARE users and developers.
I have a question regarding the unusual output I get after building the classifier... All preprocessing steps were performed using new "Hammer Suite". My input to classifier - 26 pairs of RAVENS maps (GM+WM). Everything went almost completely good except one thing... The matter is that after classifier creation I am getting very strange spatial maps. They look like four sets of bricks distributed in corners (attached)... Everything is ok with test data.
What could cause this problem?
P.S.: I'm running COMPARE on 64-bit CentOS 5.5.
I have a question regarding the unusual output I get after building the classifier... All preprocessing steps were performed using new "Hammer Suite". My input to classifier - 26 pairs of RAVENS maps (GM+WM). Everything went almost completely good except one thing... The matter is that after classifier creation I am getting very strange spatial maps. They look like four sets of bricks distributed in corners (attached)... Everything is ok with test data.
What could cause this problem?
P.S.: I'm running COMPARE on 64-bit CentOS 5.5.
Apr 12, 2011 02:04 PM | Bilwaj Gaonkar
RE: Corrupted spatial maps
This is an issue with the ANALYZE header.
Something is wrong with the header resolutions that is making COMPARE think that there are 4 images instead of one.
I think classification accuracy should not be affected by this
Something is wrong with the header resolutions that is making COMPARE think that there are 4 images instead of one.
I think classification accuracy should not be affected by this