What’s New#

v2.0.0 (2026-06-16)#

Highlight#

  • ship notebooks inside the package and fetch demo data on demand from Zenodo

  • unify the command line tools into a single minian CLI with notebooks and data subcommands

  • ~12x faster motion estimation via dask task de-duplication and a cheaper registration/warp kernel

  • consolidate video I/O into minian.io

  • rebuild the documentation toolchain (modern Sphinx, MyST, pydata-sphinx-theme) with ReadTheDocs previews

Feat#

  • bundle the pipeline and cross-registration notebooks in the package, copied out with minian notebooks copy

  • fetch demo datasets from Zenodo on first run, cached and checksum-verified, managed with minian data

  • add a single minian CLI replacing minian-data and minian-notebooks

Perf#

  • de-duplicate dask tasks and use a cheaper registration/warp kernel in motion estimation (~12x faster)

Refactor#

  • move video I/O to minian.io; importing those names from minian.utilities/minian.visualization now emits a deprecation warning

Breaking changes#

  • require Python >=3.10

  • install from PyPI or conda-forge instead of cloning the repo; notebooks and demo data are no longer in the repository tree

  • replace the minian-data/minian-notebooks commands with the single minian command

  • remove requirements/*.txt; dependencies and extras live in pyproject.toml

Docs#

  • modern Sphinx/MyST toolchain installed via a doc extra; ReadTheDocs PR previews and versioned builds

v1.3.0 (2026-06-16)#

Highlight#

  • modernize the dependency stack and support Python >=3.10

  • publish to PyPI via Trusted Publishing and register the minian project

  • remove the pymetis dependency in favor of a k-d tree spatial partitioner

Feat#

  • require FFmpeg/ffprobe before video I/O, with a clear error if missing

Fix#

  • update update_meta()

Build#

  • real linting and formatting with ruff

  • dependency cooldowns and an automated weekly lockfile refresh

  • present pip + conda-forge install and drop the stale environment.yml

v1.2.1 (2022-02-10)#

Fix#

  • avoid syntax error in update_spatial returns

v1.2.0 (2022-02-09)#

Feat#

  • use least square to produce proper scaling in temporal components and background terms

Fix#

  • rescale with normalizing factor when using normalize parameter in spatial and temporal update

  • fix unit id mismatch in spatial parameter exploration

v1.1.0 (2021-09-10)#

Fix#

  • pin jinja2 version to avoid doc build fail

  • use fft filter for peak-to-noise ratio computation

  • avoid conversion in xrconcat_recursive

Feat#

  • baseline fluorescence correction in temporal update with median filter

v1.0.1 (2021-05-05)#

Fix#

  • fix various typo and improve instructions in notebook

v1.0.0 (2021-05-03)#

Highlight#

  • use dask localcluster and throttling for all computations to reduce memory demands

  • add dedicated documentation site

  • add testing and continuous integration

  • release on conda-forge

v1.0.0rc1 (2021-04-30)#

Feat#

  • graph based resolving of mappings

Fix#

  • fix pipeline when subset is used

v1.0.0rc0 (2021-04-11)#

Candidate for first public release.