v4.0.0 Release Notes

New Features

Curve Fitting

Added curve fitting to the Color by Measure, Local Measurements and Histogram of Measurements windows. Users are now able to add normal, log normal and bimodal fits to their data, as well as display fit parameters. These are automatically added to the Report Generator.

Make Any Step Interruptible

All user editable MIPAR steps can be set to interrupt recipe execution in both the image processor and batch processor. To set a step as interruptible, select the recipe step and click the ‘Lightning Bolt’ button above the recipe panel.

Session Processor

Post Processor and Deep Learning Trainer have been combined into a single application called Session Processor. See new general layout.

If you have the Deep Learning Extension, you will only see an orange app named AI Session Processor in the launch bar. Changing between DL training and measurement modes is done through the switch above the “Training” panel. See new AI layout.

If you do not have the Deep Learning Extension, you will see a purple app named Session Processor in the launch bar. Layer selection editing and measurement UI has been improved. See new measurement layout..

Performance

Support for the latest RTX 3000s NVIDIA GPUs now allows for even greater acceleration of deep learning AI training and GPU computation.

Significantly improved the performance of processing large number of feature measurements, in many cases by orders of magnitude.

General UX performance on Mac has been improved, especially for systems running Apple Silicon processors.

Support Changes

• Official support for Windows 11 and macOS Monterey (12.0)
• Official support for NVIDIA GPU architecture Ampere (cc8.x) and CUDA Toolkit 11. CUDA Compute Capability 3.5 or higher now required. See GPU Computation for full support details and driver upgrade procedures.
  

!v4.0.0 uses a new version of the license manager under the hood. If after installation you face any issues with activation or software launch, please do the following:
- Go to your license manager and look for a key marked with a red “Cloned” or “Secure Storage ID Mismatch”
- Click the C2V button in that row
- Email that file to support@mipar.us and let us know you need to clear clone protection after upgrading

Improvements

• MIPAR will now download a newer version of the Bio-Formats library which patches the log4j vulnerability. This library is only used by those needing to open specialty file formats. Users currently running the older version will be prompted to update on launch. We strongly recommend updating. To update manually, download the new library version, place the .jar file in [Install Directory]/plugins and restart MIPAR.
• Significantly improved the performance of processing large number of feature measurements.

Bug Fixes

• Fixed bug where the recipe directory would not update after selecting a dynamic calibration recipe.
• 2D Roundness measurement no longer improperly changes with scale factor.
• Recipes with dynamic scale factors now load properly with GPU Acceleration on.
• File>Close now properly closes the window.
• Measurement dropdown now properly populates when editing a dynamic scale calibration
• Session Processor now prints companion area column title property .
• Fixed bug in the 3D Toolbox when measuring features in 2D slices that are disjoint in that slice.
• Fixed column header for Companion Perimeter measurement in the Image Processor when a calibration is present.
• Fixed NaN output when exporting some surfaces to VTK.
• Fixed bug with Smart Cluster erroring when the execution speed was set to 1.
• Replace with Local Normal no longer errors when a feature has no pixels within the specified radius of the feature’s centroid.
• Clearing unsaved session in Session Processor and choosing “Save”, would then asked whether to overwrite, despite never having saved.

Detection Changes

• Find Circles has changed, check parameters to ensure consistent behavior between v4.0.0 and previous versions.
• Local Orientation measurements may have changed for some configurations, check for consistent behavior between v4.0.0 and previous versions.
• Apply Deep Learning models may have changed for some configurations, check for consistent behavior between v4.0.0 and previous versions.

Measurement Changes

• 2D Roundness measurement no longer improperly changes with scale factor
• Updated calculation of Caliper Diameter (and by extension Roundness) measurements. Previously, Caliper Diameter was measured from the center of the two pixels farthest apart in the feature. Now, Caliper Diameter considers the entire pixel area, measuring from outside pixel corners around the feature. This will increase the Caliper Diameter measurement between 1 and 1.41 pixels (for 2D measurements) or between 1 and 1.73 pixels (for 3D measurements), depending on the orientation of each feature’s Caliper Diameter axis.