Microscopy Image Browser update to 2.80

Dear all,
today Microscopy Image Browser was updated to version 2.80. The update includes multiple improvements especially in training and segmentation of 2D convolutional neural networks with DeepMIB.
At the moment, Matlab code and Windows binaries are available. The Mac and Linux version will come soon.

Download link: Microscopy Image Browser, a handy tool for image management
or Menu->Help->Update

Release notes:
The standalone version of MIB comes with R2021a

  • Added HDD mode to align datasets that can not be fit into memory ( Menu->Dataset->Alignment tool->HDD )
  • Added new preference dialog ( Menu->File->Preferences )
  • Added exclusion of Stretch and Shear peaks into the automatic alignment using image features
  • Added re-ordering of annotations in the Annotation list via a popup menu
  • Added precision edit box into the Annotation panel
  • Added “Extra depth to show annotations” ( Annotations->Annotation list->Settings )
  • Added saving of models as 2D sequence MIB Matlab format ( Menu->Models->Save model as->Matlab format 2D sequence (*.model) )
  • Added Mode filter ( Menu->Image->Image filters... )
  • Added “File and Directory operations” to batch processing
  • Added plug-in for converting images between different formats ( Menu->Plugins->File Processing->ImageConverter )
  • [link] Added plug-in detection of contacts in 2D images ( Menu->Plugins->Organelle analysis->MCcalc )
  • Added shift of color channels in X and Y ( Menu->Image->Color channels->Shift channel )
  • Added check for a new version for MIB compiled for Linux
  • [ DeepMIB ] Added selection of GPU/CPU/Multi-GPU
  • [ DeepMIB ] Added GPU Info window ( Network panel->? )
  • [ DeepMIB ] Added possibility for training and prediction without preprocessing
  • [ DeepMIB ] Added parallel pre-processing
  • [ DeepMIB ] Added compatibility with models in TIF and PNG formats
  • [ DeepMIB ] Added masking
  • [ DeepMIB ] Added 19 2D augmentation operations with individual configuration settings
  • [ DeepMIB ] Added configurable preview of augmentation patches ( Train tab->Augmentation->Preview and ->Settings )
  • [ DeepMIB ] Added configurable ( Options tab->Custom training plot ) preview of augmented patches during training
  • [ DeepMIB ] Added possibility to select various activation layers
  • [ DeepMIB ] Added Transfer learning to update models to different number of classes
  • [ DeepMIB ] Added setting of mini-batch size for prediction
  • [ DeepMIB ] Added options for export of prediction scores
  • [ DeepMIB ] Added calculation of occurrence and Sorensen-Dice similarity for comparison of ground truth and generated models ( Prediction tab->Evaluate segmentation )
  • [ DeepMIB ] Added filenames and export to CSV to evaluate segmentation operation
  • [ DeepMIB ] Added export of trained models to ONNX format ( Matlab version of MIB )
  • Updated reading of NRRD files for MacOS
  • Updated export statistics to annotations to keep settings during current session
  • Updated Bio-Formats to 6.6.1
  • Updated for Matlab R2021a
  • Fixed missing the “Add to” checkbox Ctrl+F operation
  • Fixed loading of AM files on virtual machines due to encoding
  • Fixed preview behavior for image filters in 3D
  • Fixed 3d backup operation when number of 3D dataset == 0
  • Fixed vertical flip of MRC datasets
  • Replaced quantile function with a custom code
  • Bug fixes for interpolation, crop
  • [ DeepMIB ] Fixed loading of config on various OS
  • [ DeepMIB ] Updated training progress plot and added configuration parameters
  • [ DeepMIB ] Improved performance of image preprocessing for AmiraMesh

Bug fixes,
version 2.802

  • fix of crash in compiled application when starting color chooser
  • fix when select directories dialog sends main MIB window to top
  • fix of ExternalDirectories in preferences initialization