Skip to content

Current Release Notes

This page lists the current and potentially beta-version release notes for Microscopy Image Browser (MIB), detailing new features, improvements, and fixes across versions.
For the latest updates, visit MIB website or check the Current release notes. For the downloads visit MIB website for the development history click here.

NEXT RELEASE (in development)
  • OME-Zarr pyramids: every coarser level is now exactly half the size of the level above, so all levels line up with the full-resolution image, also in other OME-Zarr viewers. The same data gives the same pyramid whether it is converted to BigData, saved as OME-Zarr, or written by the Image converter. Pyramids written by earlier versions can differ slightly at the coarser levels when the image width or height is an odd number.
  • Image converter, OME-Zarr output:
    • faster; with Parallel processing the workers share the conversion
    • an interrupted or cancelled conversion can be resumed with the same settings
    • TIF images too large for the memory are converted in parts
    • the memory needed is checked before the conversion starts
    • a failed conversion is reported as failed, and MIB warns when an incomplete dataset is opened
    • the output directory must be empty or hold a previous Zarr dataset
  • DeepMIB, StarDist: a new architecture of the 2D Instance workflow for round and oval objects such as nuclei, round cells or vesicles, also when they are packed densely. Fine-tune the pretrained 2D_versatile_fluo model (downloaded on first use) or train a new network; no Python is needed, also in the standalone version. Includes a test of whether the objects suit StarDist and tuning of the detection thresholds after training. See 2D Instance Segmentation
  • DeepMIB, evaluation of 2D Instance predictions: Evaluate segmentation matches the predicted objects with the ground-truth objects and reports precision, recall, F1 and panoptic quality over a range of IoU thresholds, per image and for the whole set; export to MATLAB, Excel or CSV. See Evaluating the predictions
  • DeepMIB, OME-Zarr images: OME-Zarr images (v2 and v3) are predicted directly, at a chosen pyramid level. With the Zarr3 (BigData model) output, the result opens as an editable BigData model
  • BigData instance model: a new 16383-material model for BigData datasets, one index per object; 63-material BigData models can be converted to it, into a new model store
  • Close model: new entry of the buffer context menu in the Datasets panel; closes the model and keeps the dataset open, Ctrl+Z brings the model back
  • Chunk dataset: a downsampled pyramid level of a BigData or OME-Zarr dataset can be chunked; a model segmented on these tiles is fused back into the matching level of the BigData model
  • Stitching: the inspector handles mosaics of Z-stacks; the Fix Z table lists the section or slice boundaries, each opened in the Fix Z boundary view
  • Resize dataset can be cancelled; downloads show their progress
  • Fixes:
    • MATLAB no longer crashes at exit after ZARR files were opened in a session that also used Computer Vision Toolbox features
    • Image converter: OME-Zarr conversion of AmiraMesh volumes
    • DeepMIB: Export network
    • the installation directory of MIB on macOS is detected correctly
RELEASE 2026.09 / 03.06.2025 (MIB3 release)
  • Release

Back to MIB | Getting started