v4.0.4:
  Users:
    - Recover size validation in refinement. Remove deploy log messages from convert.
    - Fix motioncorr gain file location error.
v4.0.3: deploy test.
v4.0.2:
  Users:
    - RelionSubtomograms no longer have an index in the path. Just the file name. This allows for correct visualization
      of the subtomograms in the metadata viewers.
    - Analyze results now displays an auxiliar window with display options for the protocols initial volume,
      3d classification, and subtomos refine.
    - Improved Relion version detection by checking both the path and the conda environment name, ensuring
      either ends with relion-5.0.
    - The initial protocol model now supports multiple classes in Scipion.
    - Fix bug in pseudoSubtomograms2Star for 3D particles.
    - Start using the new parameter for PointerParams in forms when needed.
    - Deprecate the protocol tomo reconstruction.
   Developers:
    - The initial model protocol tests were expanded to cover multiclass and 3D particle cases.
v4.0.1:
  Users:
    - Ensure that the pseudo-subtomograms of the initial model protocol have even dimensions.
    - Post-process: the FSC generated contain the path to the postprocess.star file.
    - CTF refinement and bayesian polishing: form adapted to admit FSC as input and get the postprocess.star
      file from it.
  Developers:
    - Fix in the convert transformation matrix for negative rotation angles.
    - Add binThreads and needsGpu to the protocols.
    - Replace the setup.py + requirements.txt files by the new pyproject.toml.
v4.0.0:
  Users:
    - Relion 5.0.
    - Adapted to scipion-em-relion deprecation of relion3.
    - Add warning preventing to set the box size and the cropped box size to the same value as it may cause an
      internal error in relion (Thanks to DSheperd).
  Developers:
    - Fix inspect in buildbot: Tkagg not set in wizard module.
    - Protocols CTF refine and Bayesian polising get the postprocess.star the introduced post-processed volume, as an
      extended attribute.
    - Fix convert for 3D particles in Relion 5.
    - Improve the validations in protocols autorefine and classify 3d.
v3.4.0:
  Note: this was supposed to be 3.3.0, but was published as 3.4.0
  Users:
    - After having performed a refine or classification of particles, fields _rlnLogLikeliContribution,
      _rlnMaxValueProbDistribution and _rlnNrOfSignificantSamples are registered, so they can be plotted versus the
      no. particles using the data viewer and used as quality control metrics.
   - Add extra params to some more protocols apart from refine subtomograms.
   - Fix some initial MPI/Threads combinations (Thanks to Bram Koster).
   - Rename the protocols the same as in native Relion.
   - Enhancing the helps and use of RelionTomo4 (Thanks to J. L. Vilas).
   - Validation conditions have been release and checked (Thanks to J. L. Vilas).
  Developers:
    - Tests rewritten using the test dataset relion40_sta_tutorial_data, using the Test Centralization Layer (TCL), and
      decoupled in independent tests.
    - Remove BETA status
v3.2.2:
 Users:
  - Hotfix: the protocol visualization on the left panel was fixed.
  - Call the tlt file generation method considering the excluded views.
  - Fix the 3d classification Scipion Classes objects generation (previous bad particle assignment).
 Developers:
  - Fix: the coordinates generated in the convert input step of the prepare data protocol are now float instead of
    integer.
v3.2.1:
  - Installation requirements better managed.
  - Fix: Relion output now copies the acquisition.
  - Fix: The sampling rate displayed for the initial volume was incorrect.
  - Fix: FSC object stored unused data for the FSC viewer. That data was removed.
  - Tests updated.
v3.2.0:
  - Wizard to apply operation in relion particles
  - 3D Classification defines its possible outputs
  - An initio fixed (not using optimization set)
  - Other FSCs are plotted in the refine and postprocess protocols
v3.1.6:
  - 3D Classification more completed/standard
  - 3D classes can be fed to mapback or others
  - Refine and 3D classification use and register optimisation_set.star files

v3.1.5:
  - Prepare more robust and optimized to subsets of coordinates and Tilt series. Still ctf could be added to the ecuation.

v3.1.4:
  - GroupId is not lost through reliontomo processing
  - Fix shifts conversion in subsets: Scipion-Relion and viceversa
  - Export pseudo to coordinates removed completely
  - Apply operation works on non extracted pseudo subtomograms
  - Dataviewer shows Pseudosubtomograms better.
  - Import coordinates from star: fix scheduling (indirect pointer to precedents)
  - Prepare data:
      Fix scheduling (indirect pointer to ts from fiducials)
      Shifts in z and x are now taken from tomogram origin and into imod files
      Tilt series input is now mandatory until we manage to get the right one though the other inputs.
  - Fix refine: uses now the latest star file and priors are produced without a subset

v3.1.3:
  - Fix: Use one of the halves map in auto-refine if it exists
  - Fix: LandMark model generation in prepare when a join coordinates has happened. Sorting issue.
  - Reconstruct does not generate PseudoSubtomograms.
  - Option to "consider angles as priors" in refine and 3d classification
  - Takes into account tomogram shift (X,  Z) in prepare
v3.1.2:
  - Fix: Fiducial size in prepare is now correct (in Å)
  - Make pseudo subtomograms: float 16 is False by default and hidden. Prevents manual subsets.
v3.1.0:
  - First plugin release

