Releases: mianalysis/mia
Releases · mianalysis/mia
Version 1.7.2
Version 1.7.1
New modules
- ApplySegmentAnything creates objects using Meta's Segment Anything and the SAMJ plugin. This takes centroids of specified input objects as the point locations for segmenting objects.
- MeasureDistancesToNeightbours calculates the mean, min, max and standard deviation distances to N neighbouring objects in either the same object class or another class. These can be limited at a maximum distance.
New module features
- ImageSaver can now export images as file sequences. This includes the option to break hyperstacks down by each dimension, or keep some combined. Filename indices can be either single valued (e.g. _0001) or dimension based (e.g. _C0001_Z0002_T0001)
- ApplyDeepImageJModel can now process models that output single image classifications (rather than just pixel classification). In these cases, classifications and probabilities of being in that class are added as measurements to the relevant input image.
- Updated MeasureObjectCurvature to offer optional calculation in 3D. When in 3D only absolute curvature values are exported.
- ManuallyIdentifyObjects will show all selected objects as overlays, rather than just most recently selected object.
Other
- Updated SAMJ (Segment Anything) integration for ManuallyIdentifyObjects to be compatible with latest version of SAMJ.
- More errors and warnings will now be written to Excel log sheet
- "About" documentation page will now show version of current MIA.
- General bugfixes
Version 1.7.0
New modules
- TileStack module converts a large image into a tiled stack. This also works for hyperstacks and be set to place tiles along either of the C,Z or T axes, even if these axes already have dimensions larger than 1. Includes optional tile overlap.
- StitchTiles module will reverse tiling created using above module. Applies linear interpolation between tiles when overlap is specified. These two modules are useful when dealing with large images in DeepImageJ for example.
New module features
- Updated ObjectMeasurementCalculator to include the same options as ImageMeasurementCalculator. These modules now use a common abstract module from which they inherit much of their functionality.
- SetLookupTable can now also apply any LUTs available to ImageJ. Added the option to invert a LUT prior to applying it.
- ShowImage has additional composite display modes (invert, max and min), which can be used to display inverted (on a white background) LUTs without generating a totally white image.
Other
- Moved "Measurement" class to a slightly different location.
- General bugfixes
Version 1.6.12
Other
- Bugfixes for ManuallyIdentifyObjects
Version 1.6.11
Other
- Bugfix for batch mode file loading
Version 1.6.10
New module features
- Segment Anything implementation in ManuallyIdentifyObjects no longer allows encodings to be stored and retrieved as this used too much memory. Instead, when running in batch mode, the next image can be preinitialised, so it will load almost instantly when run properly. This requires running the workflow up to the ManuallyIdentifyObjects module in the background for the next image, so any modules displaying images will cause those images to appear.
- SetLookupTable now allows LUTs to be set based on the emitted wavelength. This uses the CIE colour matching algorithm.
Other
- Memory logging records elapsed time for each module as well as the full analysis run.
- General bug fixes
Version 1.6.9
New plugin features
- When running in batch mode, files should execute in folder order (that is, alphabetically per folder).
New module features
-
ManuallyIdentifyObjects can now implement extensions. These extensions can run code prior to the image being displayed and can add elements to the object selector control window.
-
Added support for Meta's "Segment Anything" network via SAMJ. With this, models can be automatically installed and used when running ManuallyIdentifyObjects. Selecting a region of the image will extract the corresponding object ROI using Segment Anything.
-
ManuallyIdentifyObjects gains the ability to automatically add regions to the object list when they're selected. This removes the need to continually click "Add as new object". Similarly, it can be set to add objects automatically to the existing object specified by its ID number.
Other
- General bug fixes
Version 1.6.7
New module features
- "Manually detect objects" now has an additional tool which allows the user to select any unlabelled regions within a selected ROI. To do this, the user must first select a region (as a ROI), then go to "Tools > Select unlabelled space in region". Any unlabelled pixels in that space will be selected in a new ROI, which can be added as a new object.
Version 1.6.8
Other
- Bugfix for object interpolation when using ManuallyDetectObjects
Version 1.6.6
New module features
- Running "MIA (headless)" while the macro recorder is enabled should now output the correct command. Note: the recorder will also capture a run command without parameters, which still needs to be ignored.
- Modules using object edges ("Create distance bands", "Relate many to one", "Relate many to many" and "Relate objects") can now ignore edges coincident with the image edge. When this setting is enabled, only surfaces within the image region will be considered.
Other
- Bugfixes related to modules that expand objects. Previously objects expanding beyond the image edge would offset the output objects.
- Other general bugfixes.