Changelog

This project attempts to use semantic versioning whereever possible.

[1.2.1] - 2024-01-15

Added

pcg_skel.services.get_meshwork_from_client can generate a meshwork from the information in the skeleton service.

[1.0.0] - 2021-07-02

Added

Support for metadata and meta-metadata parameters in meshparty. Now the choices made about refining a skeleton and skeletonizing a neuron will be stored in the final skeleton, enabling better control of data provenance.

Changed

PCGSkel now requires the use of caveclient instead of annotationframeworkclient as part of a general update to our suite of tools. Please install it via pip install caveclient and update your scripts accordingly, using caveclient.CAVEclient instead of annotationframeworkclient.FrameworkClient. The old client is now depricated, although no further work will go into keeping it functional.

[0.0.9] - 2021-05-06

Added

Get id under point.

[0.0.8] - 2021-05-05

Added

Synapse querying for meshworks can use the live query option.
Setting refine="chunk" in skeleton or mesh creation returns vertices in chunk indices, but only after any potential soma collapse in euclidean coordinates.

Fixed

Utility function to convert nm to chunk index had plus that should have been a minus.

[0.0.7] - 2021-04-29

Fixed

Small behind the scenes fix to accomidate changes in AnnotationFrameworkClient API.

[0.0.6] - 2021-04-10

Added

Added chunk_index_mesh function to download a level 2 object graph with chunk index vertices.

Changed

segmentation_fallback defaults to False, because this method can be very slow.

Fixed

Force bbox dimensions to be integers, which can happen with non-integer resolutions in cloudvolume.
n_parallel = None now defaults to parallel = 1 in cloudvolume, avoiding an error.

[0.0.5] - 2021-03-03

Fixed

The segmentation fallback method has been made more robust to level 2 chunk ids with very small supervoxels.
nan_rounds set to None is more correctly handled and the warning actually checks to see if any nans are left.

[0.0.4] - 2021-02-24

Added

pcg_meshwork takes the same save_to_cache argument as pcg_skeleton.

Changed

Voxel resolution is always inferred from the mip-0 resolution of the segmentation cloudvolume. Note that this means that if the segmentation mip-0 resolution is not the same as the normal working resolution, root_point_resolution must be explicitly set.
The package explicitly uses the authentication token set in the client to instantiate cloudvolume objects, enabling it to work with non-default tokens.
pcg_meshwork now sets the point column for synapses for better integration with meshwork features.

Fixed

Fixed bug in get_level2_synapses prevented post=True, pre=False situation from working correctly.
Improved default parameters in call cases where a CloudVolume instance is initialized.

[0.0.3] - 2021-02-13

Changed

Choose your mip level when falling back to segmentation. This saves a bit of time on the download, and if the level-2 object has been downsampled out of existence at this mip level, it will rerun at mip-0 where it must exist.
More aggressive handling of segmentation memory use. Requires cloudvolume renumber mode.

[0.0.2] - 2021-02-13

Added

Localizing level 2 ids via segmentation if the mesh does not exist. There are both expected and unexpected an unexpected reasons for mesh fragments not to exist. Unfortunately, this situation seems to crop up for virtually every neuron. However, all level 2 ids must be associated with at least one supervoxel, so we can use the segmentation itself to do the same operation. This is a bit slower, since we have to download the whole chunk, but it's a robust fallback that can be used when no mesh is found. This is now used by default, but can be turned off with segmentation_fallback=False on any of the functions that use level 2 positions.
Improved parallelization. To speed up the slow segmentation fallback, parallelization occurs in more processing steps.

[0.0.1] - 2021-02-11

Initial release