New plugin design, models and automatic segmentation

Version 0.5

This release redesigns the user interface and provides it as napari plugins. It also provides updated models that are now hosted on https://bioimage.io/#/ and several other improvements and new features.

We are working on updating the documentation to reflect this new release ASAP and will also create a 1.0.0 release, focusing on minor improvements and bugfixes, soon to mark the tool as stable.

What's Changed

• Sync dev and master by @constantinpape in #328
• Prototype for annotator plugins by @constantinpape in #304
• Add Lucchi amg results by @anwai98 in #340
• Add SNEMI amg results by @anwai98 in #339
• Add MitoEM results by @anwai98 in #343
• Add Covid-IF results by @anwai98 in #345
• Update Covid-IF Results (for cyto2) by @anwai98 in #346
• Add LiveCELL Vanilla Results by @anwai98 in #348
• Add livecell cellpose results by @anwai98 in #349
• Add scripts for plotting results by @constantinpape in #313
• Refactor livecell evaluation by @anwai98 in #344
• Add EM generalist evaluation by @anwai98 in #320
• Add partial finetuning submission script by @anwai98 in #352
• Add n_objects experiment by @anwai98 in #353
• Implement instance segmentation merging across 3d by @constantinpape in #351
• Add evaluation scripts for experiments by @anwai98 in #354
• Saving the grid search parameters by @anwai98 in #362
• Overwrite existing keybindings to avoid error when re-opening plugin widgets by @GenevieveBuckley in #366
• Add description for plotting scripts by @anwai98 in #365
• Update Mito Nuc Finetuning by @anwai98 in #364
• Add plot scripts for additional experiments by @anwai98 in #360
• Add LM Specialist Evaluation by @anwai98 in #310
• Refactor EM generalist evaluations by @anwai98 in #368
• Minor fix to preprocessing LM datasets by @anwai98 in #374
• Minor fix to saving grid search params by @anwai98 in #373
• Fix test failure due to change in torchvision by @constantinpape in #375
• Refactor plotting scripts by @anwai98 in #372
• Refactor all evaluation scripts by @anwai98 in #376
• Update function signatures by @anwai98 in #377
• Add override to folder paths for evaluation by @anwai98 in #378
• Add description to more plotting scripts by @anwai98 in #367
• Add M1 MacOS runner to github actions jobs by @GenevieveBuckley in #370
• Update LM generalist training (with new decoder setup) by @anwai98 in #379
• Add NeurIPS CellSeg Specialist by @anwai98 in #363
• Refactor CellPose scripts by @anwai98 in #347
• Add plotting scripts for evaluation by @anwai98 in #387
• Minor update to finetuning scripts by @anwai98 in #386
• Minor fix to documentation by @anwai98 in #399
• Update default parameter for foreground smoothing by @anwai98 in #396
• Update evaluation plot scripts by @anwai98 in #401
• Update run_all_finetuning.py - for flexible finetuning models by @anwai98 in #395
• Minor refactor to label transform in LM generalist by @anwai98 in #394
• Add inference for cellpose 3.0 by @anwai98 in #393
• Restore EM boundaries finetuning by @anwai98 in #403
• Update JointSamTrainer's save_checkpoint by @anwai98 in #404
• Shorte widget names and suppress mobile-sam import warnings by @constantinpape in #409
• Batch point seg by @constantinpape in #410
• Minor fix to inference function for instance segmentation by @anwai98 in #411
• Add resource efficient finetuning experiments by @anwai98 in #391
• Add volumetric instance segmentation scripts by @anwai98 in #402
• Compute best grid search parameters by @anwai98 in #413
• Implement new segmentation plugin by @constantinpape in #305
• 331 add options for tiling in the embdding widget by @lufre1 in #419
• Update projection modes by @constantinpape in #418
• Update CLI to enable loading a finetuned model by @constantinpape in #427
• Fix initialization with segmentation result by @constantinpape in #431
• Enable tile based segmentation for more options and add more tests by @constantinpape in #432
• Enable specifying the device in the CLI by @constantinpape in #433
• Parallelize commit operations by @constantinpape in #436
• Add input logits logic to iterative prompting evaluation by @anwai98 in #438
• Add exception message for no objects found by @anwai98 in #443
• Clear annotation in slice or frame only by @constantinpape in #447
• Implement commit restricted to actually segmented slices for 3d annotator by @constantinpape in #448
• Implement commit to file by @constantinpape in #451
• Update model comparison to work with prompt gen syntax by @constantinpape in #453
• Fix issues in loading segmentation result and image series annotator by @constantinpape in #456
• Update image_series_annotator - for flexible inputs by @anwai98 in #454
• Update evaluation scripts for custom experiments by @anwai98 in #442
• Update inference and evaluation top-level scripts (for LIVECell) by @anwai98 in #439
• Amg artifacts by @constantinpape in #458
• Include min_size in multi dim evaluation by @anwai98 in #459
• Unify SAM model loading by @constantinpape in #460
• Update supplementary experiments by @anwai98 in #420
• Update current metric saving in joint trainer by @anwai98 in #461
• Add convenience function for training and add user study scripts by @constantinpape in #457
• Fix issue with segmentation for non-square images...

0.4.1

What's Changed

• Minor doc fix by @anwai98 in #329
• Update finetuning example by @anwai98 in #338
• Minor fix to documentation by @anwai98 in #357
• Explain import of submodules by @constantinpape in #359
• Fix amg for the image series annotator by @constantinpape in #398

Full Changelog: 0.4.0...0.4.1

0.4.0

What's Changed

• More comprehensive .gitignore file at repository root by @GenevieveBuckley in #222
• CI testing - automatic code coverage reports by @GenevieveBuckley in #214
• Simple GUI test for CI by @GenevieveBuckley in #223
• Improve test train by @constantinpape in #233
• Update CI on master by @constantinpape in #232
• Make numba optional by @constantinpape in #238
• Add documentation for how to run micro-sam via BAND by @constantinpape in #229
• Add LM Generalist Finetuning by @anwai98 in #228
• Fixes for test_training.py failures by @GenevieveBuckley in #245
• Add EM Generalist Finetuning (2) by @anwai98 in #247
• Developer documentation by @GenevieveBuckley in #243
• Documenation build script - add check for missing docs files by @GenevieveBuckley in #250
• Merge master into dev by @constantinpape in #252
• Add support for batched inference with prompts by @constantinpape in #253
• Implement singleton state and use it in the 2d annotator by @constantinpape in #240
• Add Histopathology Generalist Training by @anwai98 in #248
• Fix Bounding Box Sorting by @anwai98 in #257
• Refactor Iterative Prompting Evaluation by @anwai98 in #254
• Add 3d segmentation functionality by @constantinpape in #255
• Fix EM Generalist Raw Transform by @anwai98 in #259
• Update Evaluation - Iterative Prompting by @anwai98 in #260
• Embedding widget for napari plugin by @GenevieveBuckley in #235
• Use a consistent cache directory by @GenevieveBuckley in #268
• Fixup README 404s by @claydugo in #270
• Fix typo in function name _segment_slice_widget by @GenevieveBuckley in #273
• Update Description - LiveCELL Inference by @anwai98 in #271
• Use qtpy by @constantinpape in #275
• Experiments with Mask Inputs Probability by @anwai98 in #265
• Merge dev into master by @constantinpape in #278
• Keep master and dev in sync by @constantinpape in #279
• napari widget to set cache directory (for model weights and sample data downloads) by @GenevieveBuckley in #277
• Run github actiosns CI tests on Windows, Mac and Linux by @GenevieveBuckley in #267
• Improve device handling by @constantinpape in #251
• Increase test coverage by @constantinpape in #280
• Fixup typo by @claydugo in #281
• Fix Plugin Link by @anwai98 in #282
• Download models with pooch by @GenevieveBuckley in #276
• Add xxh128 version of hashes for model by @hmaarrfk in #284
• Update Rescaling Inputs by @anwai98 in #286
• Add Joint Training by @anwai98 in #288
• Fix load_checkpoint by @anwai98 in #290
• Update LiveCELL Specialist Fine-tuning by @anwai98 in #291
• Update livecell evaluation scripts by @constantinpape in #296
• New instance segmentation by @constantinpape in #297
• Update finetune_hela.py by @Mobinapournemat in #294
• Fix Resizing Input Prompts by @anwai98 in #299
• Minor fix to Instance Segmentation by @anwai98 in #302
• Simplify training implementations by @constantinpape in #301
• Make grid search functionality more flexible by @constantinpape in #300
• Update livecell finetuning by @anwai98 in #303
• Update LM Generalists by @anwai98 in #292
• Update LM Generalist Dist. Trafo. by @anwai98 in #306
• Update README.md - Minor doc fix by @anwai98 in #307
• Add EM Mito-Nuc Generalist by @anwai98 in #295
• Implement grid search for livecell training and update instance seg grid search by @constantinpape in #308
• Minor update to EM finetuning by @anwai98 in #309
• Minor typo in em mito nuc by @anwai98 in #311
• Add EM Boundaries Generalist by @anwai98 in #312
• Minor updates to finetuning desc. by @anwai98 in #314
• Update LiveCELL inference scripts by @anwai98 in #315
• Fix instance segmentation checkpoint loading by @anwai98 in #318
• Add livecell specialist results by @anwai98 in #317
• Add livecell (lm) generalist results by @anwai98 in #319
• Minor fix to livecell eval. scripts by @anwai98 in #316
• Minor fix to experiment path convention by @anwai98 in #323
• Refactor files in scripts by @anwai98 in #321
• Update mitonet results by @anwai98 in #325
• Update to new finetuned models by @constantinpape in #326
• Changes for new release by @constantinpape in #327

New Contributors

• @claydugo made their first contribution in #270
• @hmaarrfk made their first contribution in #284
• @Mobinapournemat made their first contribution in #294

Full Changelog: 0.3.0...0.4.0

0.3.0post1

Fix tests for conda forge.

0.3.0pos2

Release for 0.3.0 binaries.

0.3.0

This release introduces two new features:

• Segmentation from shapes other than bounding boxes (ellipses and polygons)
• Automatic instance segmentation for 3d

In addition we have refactored and improved the training code, which speeds up model finetuning.

See below for a more detailed changelog.

What's Changed

• More doc updates by @constantinpape in #162
• Using predictor device when setting image embedding by @JoOkuma in #163
• micro-sam napari plugin by @GenevieveBuckley in #165
• Fix YAML formatting errors so Github actions will run by @GenevieveBuckley in #166
• Implement 3d auto segmentation by @constantinpape in #169
• fixed annotator_3d missing embedding path by @tc2fh in #170
• Supporting Apple Silicon devices by @GenevieveBuckley in #178
• Add MPS Pytorch support for batched_mask_to_box function by @GenevieveBuckley in #180
• Add support for MPS device by @constantinpape in #176
• Clear torch memory cache in tests by @GenevieveBuckley in #179
• Implement runtime benchmarks by @constantinpape in #182
• Add initial support for vit tiny model from mobileSAM by @constantinpape in #174
• Clean up by @constantinpape in #184
• Trying to improve performance for MPS by @constantinpape in #183
• Add docstring about how to run benchmarks by @GenevieveBuckley in #187
• Implement RLE computation with numba by @constantinpape in #190
• Refactoring Iterative Training Scheme by @anwai98 in #192
• Implement prototype for segmentation of complex shapes by @constantinpape in #196
• Fix CLI for embedding precomputation by @constantinpape in #209
• Refactor Prompt Generators by @anwai98 in #210
• Fix tuple annotations by @constantinpape in #213
• Fix prompt gens by @constantinpape in #211
• Change keybinding for volume segmentation (remove clash with napari layer visibility keybinding) by @GenevieveBuckley in #207
• Next release by @constantinpape in #172
• Implement integration test for training by @constantinpape in #198
• Minor Fix - Livecell Patch Shape by @anwai98 in #216
• Add imagecodecs dependency by @constantinpape in #217
• Update Casting of Instance Estimations by @anwai98 in #218
• Update Documentation by @anwai98 in #220
• Fix Imports in Livecell Inference by @anwai98 in #219
• Fix Box Prompting in Inference by @anwai98 in #221
• Merge fixes into master by @constantinpape in #224
• Histopatho model by @constantinpape in #193
• Add prediction with box prompts to training integration tests by @constantinpape in #225
• Bump version for release 0.3.0 by @constantinpape in #226

New Contributors

• @tc2fh made their first contribution in #170

Full Changelog: 0.2.2...0.3.0

0.2.2

What's Changed

• Refactor state precomputation by @constantinpape in #155
• Documentation updates by @constantinpape in #152
• Fixing typo 'f.atrs' to 'f.attrs' by @JoOkuma in #156
• Serialize amg state to the cpu so that it can be loaded without a GPU by @constantinpape in #157
• Make embedding file checks more robust by @constantinpape in #158
• Precompute embeddings in 2d example; set default box extension to 0.0… by @constantinpape in #159

New Contributors

• @JoOkuma made their first contribution in #156

Full Changelog: 0.2.1...0.2.2

0.2.1post1

Release for building binaries.

0.2.1

What's Changed

• Remove opencv dependency by @constantinpape in #147
• Fix amg state precomputation for image series annotator by @constantinpape in #148
• Support RGB data in micro_sam.evaluation.model_comparison by @constantinpape in #149
• Finetune example by @constantinpape in #150

Full Changelog: 0.2.0...0.2.1

Post release for binaries

Merge pull request #146 from computational-cell-analytics/post-release

Bump version and update texts for installer