24.08.0

• Docker Image scalableminds/webknossos:24.08.0 🐳
  Run it via export DOCKER_TAG=="24.08.0" && ./start-docker.sh.
• Changelog
• Commits

Highlights

• The AI-based Quick Select can now be run on multiple sections at once. This can be configured in the tool settings. Also, the underlying model now uses Segment Anything 2. #7965
• Metadata entries can now be added to datasets and folders. The metadata can be viewed and edited in the dashboard in the right details tab. #7886
• The AI-based Quick Select can now be triggered with a single click. Drawing a rectangle is still supported. #7993
• Added the option to resume an unfinished upload even after browser restarts. #7981
• Added a feature to register all segments for a given bounding box at once via the context menu of the bounding box. #7979

Added

• WEBKNOSSOS now automatically searches in subfolder / sub-collection identifiers for valid datasets in case a provided link to a remote dataset does not directly point to a dataset. #7912
• Added the option to move a bounding box via dragging while pressing ctrl / meta. #7892
• Added route /import?url=<url_to_datasource> to automatically import and view remote datasets. #7844
• Added that newly created, modified and clicked on bounding boxes are now highlighted and scrolled into view, while the bounding box tool is active. #7935
• The configured unit in the dataset upload view is now passed to the convert_to_wkw worker job. #7970
• Added option to expand or collapse all subgroups of a segment group in the segments tab. #7911
• The context menu that is opened upon right-clicking a segment in the dataview port now contains the segment's name. #7920
• Upgraded backend dependencies for improved performance and stability. #7922
• Added Support for streaming datasets via Zarr version 3. #7941
• It is now saved whether segment groups are collapsed or expanded, so this information doesn't get lost e.g. upon page reload. #7928
• It is now saved whether skeleton groups are collapsed or expanded. This information is also persisted to NML output. #7939
• The context menu entry "Focus in Segment List" expands all necessary segment groups in the segments tab to show the highlighted segment. #7950
• In the proofreading mode, you can enable/disable that only the active segment and the hovered segment are rendered. #7654
• Upgraded s3 client for improved performance when loading remote datasets. #7936
• The performance of the bounding box tab was improved. #7974
• Added support for reading zstd-compressed zarr2 datasets #7964
• The alignment job is in a separate tab of the "AI Tools" now. The "Align Sections" AI job now supports including manually created matches between adjacent section given as skeletons. #7967
• Added api.tracing.createNode(position, options) to the front-end API. #7998
• Added links in the workflow report for skipped tasks to the corresponding workflow view. #8006
• Upgraded backend dependencies for improved performance and stability, including a memory leak fix for FossilDB. #8014

Changed

• Replaced skeleton tab component with antd's <Tree />component. Added support for selecting tree ranges with SHIFT. #7819
• The warning about a mismatch between the scale of a pre-computed mesh and the dataset scale's factor now also considers all supported mags of the active segmentation layer. This reduces the false posive rate regarding this warning. #7921
• It is no longer allowed to edit annotations of other organizations, even if they are set to public and to others-may-edit. #7923
• When proofreading segmentations, the user can now interact with super-voxels directly in the data viewports. Additionally, proofreading is significantly faster because the segmentation data doesn't have to be re-downloaded after each merge/split operation. #7654
• Changed internal data model changing an organization's name to id and its displayName to name. The previously existing id was removed. #7386
• Because of the way our models are trained, AI analysis and training is disabled for 2D and ND datasets, as well as for color layers with data type uInt24. #7957
• The overall performance was improved (especially for the segments tab). #7958
• The performance for the skeleton tab was improved. #7989
• Upgraded ant icons to version 5.4. #8007
• Increased maximum depth for ai-based quick select from 5 to 16. #8021
• Zarr-streaming now uses the new unit-aware voxel size. This means that wk-libs client v0.14.25 or newer is needed for dataset zarr streaming. #8012
• Updated React to version 17. #7765

Fixed

• Fixed a bug that allowed the default newly created bounding box to appear outside the dataset. In case the whole bounding box would be outside it is created regardless. #7892
• Fixed a rare bug that could cause hanging dataset uploads. #7932
• Fixed that comments of the active tree were not scrolled into view in some cases when switching to the comments tab. 8022
• Fixed that trashcan icons to remove layers during remote dataset upload were floating above the navbar. #7954
• Fixed that the flood-filling action was available in the context menu although an editable mapping is active. Additionally volume related actions were removed from the context menu if only a skeleton layer is visible. #7975
• Fixed that activating the skeleton tab would always change the active position to the active node. #7958
• Made the newly added filePaths argument of the reserve upload route (see #7981) optional to be backwards compatible with wklibs. #8045
• Fixed that skeleton groups couldn't be collapsed or expanded in locked annotations. #7988
• Fixed that the dashboard didn't notify the user about new datasets in the table. #8025
• Fixed that registering segments for a bounding box did only work if the segmentation had mag 1. #8009
• Fixed uploading datasets in neuroglancer precomputed and n5 data format. #8008
• Various fixes for composing datasets with landmarks. Note that the interpretation of correspondence points was inverted for thin plate splines. #7992