Release v3.2.2

Major changes:

• Fix bug when saving several classification results (use correct LUT and image type).

Release v3.2.1

Major changes:

• Fix bug when balancing classes from macro.
• Do not show membrane options in settings when working in 3D.

Release v3.2.0

Major changes:

• Add 3D GUI compatibility, including new plugin.
• Use imglib2 for 3d DoG features.
• Update WEKA version to 3.9.0.
• Rename "homogenize" methods by more accepted "balance".
• Fix several bugs related to displaying results.

Release v3.1.2

Major changes:

• Fix java version problem with release 3.1.1.
• Increase maximum number of classes to 100 (release 3.1.1).
• Add method to enable/disable a feature by name (release 3.1.1).
• Add library methods to add training instances from a label image (release 3.1.1).
• Much code cleanup by Mohamed Ezzat from DevFactory (release 3.1.1).

Release v3.1.0

Major changes:

• Correct wrong method name for V_info thin metric.

Release v2.3.0

Major changes:

• Fix bug on float number comparisons by using Float.compare()
• Speed up classification of new images by reusing a weka instance for each pixel.
• Include new metrics V_Rand (thin) and V_info (thin) as described in our Frontiers in Neuroanatomy paper.

Release v2.2.1

Major changes:

• Fix bug to return result from macro when classified image exists.
• In the Entropy filter, convert to image to byte before calculating bins.

Release v2.2.0

Major changes:

• Make ImageScience features optional, thanks to Curtis Rueden. This way we avoid license problems.
• Propagate calibration to result image, thanks to Jan Eglinger.
• Add public methods to calculate disagreements based on Rand or Information theoretic scores

Release v2.1.8

• Fix bug reported by Martin Höhne to be able to calculate the result image from a macro after having loaded a classifier.

Release v2.1.7

Major changes:

• Clean RandError metric class to clearly separate the standard Rand index/error calculation from the foreground-restricted version that is used in the 2D EM segmentation challenge (http://brainiac2.mit.edu/isbi_challenge/).
• In the GUI, fix bug that was causing disabled buttons after an error in Plot results (pointed out by Mark Hiner).
• In the GUI, allow creating result after loading a classifier (also suggested by Mark Hiner).
• Add a method to load a classifier from an input stream (by Patrice Freydiere).