Incorrect Assignment of Class

I am working on a project in 0.2.0-m2 with a colleague and he made an object classifier, which he sent to me, along with a class file and script. It seems to be picking up all positive cells, but scrambling up which category they belong to (consistently - i.e all cells of x category are marked as y and all cells of y category are marked as x).

Any suggestions why this might be and/or potential fixes appreciated! :slight_smile:

It could be that object measurements are different/missing in the new project from the ones originally used to create the object classifier.

(Note that v0.2.0-m2 is a prerelease version from well over a year ago… I’d highly recommend updating to the latest stable release, which is currently v0.2.1. However I don’t recommend training classifiers and applying them in different versions of the software.)

2 Likes

If it is 100% consistent, could you swap the names in the exported statistics?
Alternatively, you could swap the classes with a script in m2, or even through the GUI in 0.2.2.
How exactly that would work would depend on how many classes you have, but the basic idea would be

changeToY = getCellObjects().findAll{it.getPathClass() == getPathClass("X")}
changeToX = getCellObjects().findAll{it.getPathClass() == getPathClass("Y")}

changeToY.each{it.setPathClass(getPathClass("Y"))}
changeToX.each{it.setPathClass(getPathClass("X"))}
1 Like

Update: Thanks Pete and Research_Associate - turns out it just needed a restart of the software!