v0.4.6

[0.4.6] - 2025-12-16

Added

• Added metadata section to video and video frame details.
• Added tag support for videos and video frames in the GUI.
• Introduced navigation between video details.
• Enabled video and video frames filtering.
• Added text search for videos.
• Added plugins: This is the initial version for plugins. It supports the execution of operators.
• Added cloud storage support for video frames.
• Added export for image captions.
• Added semantic search by adding perception encoder core as embeddings model.
• Added VideoSample class.

Changed

• Renamed lightly_studio.core.sample.Sample to ImageSample.

What's Changed

• Modify the video frame endpoint to support filtering by @LeonardoRosaa in #234
• Implement annotation count for video frames by @LeonardoRosaa in #242
• Implement the video fields filtering within the GUI by @LeonardoRosaa in #239
• Fixed annotations menu icon by @LeonardoRosaa in #248
• Add endpoint to retrieve video frame fields bounds by @LeonardoRosaa in #243
• Add video frame annotations filtering to GUI by @LeonardoRosaa in #244
• Add video annotations filtering to GUI by @LeonardoRosaa in #246
• Run captions e2e tests from CI by @michal-lightly in #238
• Removed dataset_id from AnnotationBaseTable by @horatiualmasan in #236
• Video Embedding: Add perception encoder vendor model by @JonasWurst in #251
• Refresh header after annotation/caption creation by @LeonardoRosaa in #254
• Update counter samples information in GUI for the video and video frames by @LeonardoRosaa in #247
• Prevent video playback when not hovering by @LeonardoRosaa in #257
• Operator Dialog by @JonasWurst in #249
• Add metadata filters for videos frames. by @LeonardoRosaa in #253
• Use tags for video samples in GUI by @horatiualmasan in #250
• Video Embedding: Download from lightly hosted s3 by @JonasWurst in #256
• VideoEmbeddingGenerator Protocol + EmbeddingManager refactor by @JonasWurst in #259
• Make Dataset generic by @lukas-lightly in #260
• Add captions e2e test - sample details page by @michal-lightly in #245
• Add metadata section for video and video frames by @LeonardoRosaa in #255
• Rename Sample to ImageSample by @lukas-lightly in #263
• Allow filter by video frame number by @LeonardoRosaa in #252
• Embedding Manager: Add Video Embed + tests by @JonasWurst in #262
• Exclude unused annotation labels from annotations counts endpoint by @LeonardoRosaa in #267
• Apply filter for video frame details navigation by @LeonardoRosaa in #266
• Add video details navigation by @LeonardoRosaa in #264
• Perception Encoder Embedding Generator: Text + Image by @JonasWurst in #265
• Split responsibilities in export samples dialog by @michal-lightly in #268
• Add e2e tests for the captions page by @michal-lightly in #261
• Make Export always available by @JonasWurst in #269
• Perception Encoder Emb Gen: add video embedding by @JonasWurst in #270
• Introduce a new endpoint to fetch annotations and support multiple samples by @LeonardoRosaa in #273
• Enable video frame view cloud support by @horatiualmasan in #276
• Add tags to video frames by @horatiualmasan in #258
• Introduce endpoint to fetch dataset details for dashboard by @mihnea-necsulescu in #272
• Update changelog for cloud video frame support by @horatiualmasan in #279
• Export for image captions from GUI by @michal-lightly in #271
• Annotation grid: implement new data structure for images by @LeonardoRosaa in #278
• Add our own log formatter by @lukas-lightly in #277
• Do not print uvicorn access log by default by @lukas-lightly in #281
• annotations: adjustment to support multiple sample types by @LeonardoRosaa in #280
• Introduce ImageSample.sample_table by @lukas-lightly in #283
• Add embedding to video dataset by @JonasWurst in #274
• Create an abstract Sample class by @lukas-lightly in #284
• annotation grid: introduce video support by @LeonardoRosaa in #282
• Video embedding model: add text search bar for videos by @JonasWurst in #287
• Redirect to proper sample type based on dataset by @mihnea-necsulescu in #289
• annotations: add a new endpoint to retrieve annotation by id by @LeonardoRosaa in #285
• Remove Duplicated slider by @JonasWurst in #293
• Pass dataset_id to embedText API endpoint by @michal-lightly in #288
• Update the header component in preparation for multiple datasets support by @mihnea-necsulescu in #291
• Introduce video frame annotation details by @LeonardoRosaa in #290
• Pass datasetId with embbedings2d endpoint request by @michal-lightly in #297
• Add VideoSample class by @lukas-lightly in #294
• Pass datasetId to list samples endpoint by @michal-lightly in #298
• Enable video frame annotations filtering by @LeonardoRosaa in #292
• Eagerly load annotation tags when listing images by @horatiualmasan in #299
• Pass dataset_id to get_dataset_hierarchy endpoint by @michal-lightly in #301
• Update the breadcrumb components in preparation for multiple datasets by @mihnea-necsulescu in #302
• Plugins Docs by @JonasWurst in #295
• Order datasets by creation date for overview by @mihnea-necsulescu in #307
• Bump version to 0.4.6 by @michal-lightly in #308

Full Changelog: v0.4.5...v0.4.6

v0.4.5

[0.4.5] - 2025-12-02

Added

• Added reading and updating of captions to the Sample class.
• Added export functionality for image datasets with captions to the python Dataset class interface.
• Added keyboard shortcut support for toggling annotation edit mode.
• Added sliders to adjust brightness and contrast for more accurate labeling.
• Added version info to the footer.

Changed

• Improved the undo functionality. Works now in more views and one can undo the creation and deletion of annotations.
• Reduced the minimum size of bounding box creation from 10px side length to 4px.
• Print server errors to the console.
• Header actions (classifier, selection, export, settings) are grouped into a menu.

Fixed

• Fixed an issue with wrongly displayed annotations grid view in the presence of classification annotation type.
• Fixed video sample type in examples in readme and docs.
• Fixed a problem with listing all items on scroll in samples and annotations grid views.
• Fixed an image size reading issue for some JPEG formats.

What's Changed

• Print server errors in the console by @michal-lightly in #203
• Add additional tests for the similarity endpoint by @lukas-lightly in #201
• Create a sample for annotation by @horatiualmasan in #210
• Add a function to convert captions dataset to coco dict by @michal-lightly in #206
• Add getter/setter methods for captions by @lukas-lightly in #213
• Fixed the annotation-details e2e test by @LeonardoRosaa in #215
• Add export for captions to python interface by @michal-lightly in #217
• Fix for issues related with hitting limit of sqlalchemy pool + TCP connections for by @ikondrat in #214
• Update frontend code to use sample_id for annotations by @horatiualmasan in #216
• Example of fixing unknown type problem by @ikondrat in #219
• Use hotkey for toggling label editing mode by @IgorSusmelj in #221
• Add filter support for the video endpoint by @LeonardoRosaa in #220
• Add slider for brightness and contrast go GUI by @IgorSusmelj in #224
• Add metadata filters for videos by @LeonardoRosaa in #222
• Reduce minimum size of box side from 10px to 4px by @IgorSusmelj in #226
• Fix video sample type in example in readme and docs by @IgorSusmelj in #227
• Introduce Logging Step 1 by @masakljun in #198
• Split caching steps in e2e tests yaml by @michal-lightly in #218
• Use dialog for export. SidePanel removal by @JonasWurst in #208
• Operator Dropdown by @JonasWurst in #205
• Show LightlyStudio version in the footer by @michal-lightly in #231
• Make undo operation work more consistenly by @IgorSusmelj in #225
• Dialog for FSC by @JonasWurst in #209
• Add captions export endpoint by @michal-lightly in #228
• Prepare folder structure for caption e2e tests by @michal-lightly in #233
• Add endpoint to retrieve video fields bounds by @LeonardoRosaa in #223
• Implement a new endpoint to count annotations for videos by @LeonardoRosaa in #232
• Skip classification annotation type in annotations grid view by @LeonardoRosaa in #237
• Header Menu by @JonasWurst in #235
• Use labelformat 0.1.9 by @michal-lightly in #241
• Bump version to 0.4.5 by @ikondrat in #240

Full Changelog: v0.4.4...v0.4.5

v0.4.4

[0.4.4] - 2025-11-26

Added

• Added video support. A video dataset can be loaded from a local folder and inspected in the GUI.
• Added video annotation support to the GUI. Video annotations can be currently loaded and exported only manually.
• Renamed Dataset.add_samples_from_path to Dataset.add_images_from_path.
• Added class balancing with a uniform or the input distribution as target. These options can be set for the AnnotationClassBalancingStrategy.
• Added download_example_dataset utility function to simplify the quickstart experience by removing the need for git clone.
• Added tag_depth parameter to Dataset.add_samples_from_path to automatically create tags from subdirectory names.
• Added labeling support for captions: Add/delete/edit captions from the GUI
• Added similarity metadata calculation to Dataset.

Changed

• Renamed the distribution field of AnnotationClassBalancingStrategy to target_distribution.
• Display multiple captions per image in the Captions view.

Fixed

• Support pyav >= v14 by removing the deprecated av.AVError import.

What's Changed

• Split image_router and rename functions by @michal-lightly in #134
• Rename frontend hooks from sample to image by @michal-lightly in #135
• Rename AnnotationClassToTarget.distribution by @lukas-lightly in #138
• Add get_root_dataset to dataset_resolver by @horatiualmasan in #140
• Add get_filtered_samples function by @michal-lightly in #141
• Use label names in target class balancing by @lukas-lightly in #139
• Add an API endpoint to list samples by @michal-lightly in #142
• Add API endpoint to provide the root dataset by @horatiualmasan in #145
• Added video grid view by @LeonardoRosaa in #118
• Sample Detail view: editable captions by @JonasWurst in #132
• Show captions by sample in caption grid view by @michal-lightly in #146
• Add API endpoint to provide the dataset hierarchy by @horatiualmasan in #147
• Add tagging info to readme and docs and changelog by @masakljun in #130
• Initial implementation of dynamic dataset hierarchy by @LeonardoRosaa in #150
• Added frames grid view by @LeonardoRosaa in #119
• Added video details view by @LeonardoRosaa in #122
• Add a demo dataset download function directly to the studio by @masakljun in #137
• Selection: Add annotation_balancing syntactic sugar by @masakljun in #144
• Remove unused caption endpoint by @michal-lightly in #149
• Use SampleFilter in ImageFilter by @michal-lightly in #151
• Caption editable in Caption GridView by @JonasWurst in #155
• Use correct matrix.python version in unittests by @MalteEbner in #154
• Refactor CaptionFiled by @JonasWurst in #158
• Caption Delete Route and Resolver by @JonasWurst in #156
• Enable multithreading for video decoding by @horatiualmasan in #153
• Add Caption Resolver and Route by @JonasWurst in #157
• Added frame details view by @LeonardoRosaa in #123
• Use separate environment variables for test scripts by @michal-lightly in #163
• Rename CaptionTable.caption_id to sample_id by @michal-lightly in #159
• Add function to get or create an arbitrary child dataset by @michal-lightly in #165
• Simplify batching logic when loading samples by @michal-lightly in #167
• Add quotes to pip install extras for ZSH compatibility by @masakljun in #169
• Remove AnnotationImageView and AnnotationDetailsView by @horatiualmasan in #161
• Add Caption Frontend by @JonasWurst in #166
• Delete Caption Frontend by @JonasWurst in #160
• Remove caption test fixture by @JonasWurst in #168
• Changelog for Caption Labeling Support by @JonasWurst in #170
• Document Class Distribution Balancing Selection by @masakljun in #162
• Update development guidelines by @michal-lightly in #173
• Add compute_similarity_metadata function by @lukas-lightly in #172
• Link CaptionTable to SampleTable by @michal-lightly in #164
• Link annotations to SampleTable by @horatiualmasan in #171
• Remove CaptionTable.dataset_id by @michal-lightly in #176
• Class balancing for an incomplete target by @lukas-lightly in #177
• Fix metadata filters for infinite grid query by @MalteEbner in #181
• Create a child dataset for annotations by @horatiualmasan in #175
• Add tests for input and uniform distributions by @masakljun in #178
• Add compute_similarity_metadata to Dataset by @lukas-lightly in #180
• Support pyav versions >= v14 by @MalteEbner in #182
• add example and launch instruction to index video by @ikondrat in #187
• Add example script to load video annotations by @michal-lightly in #185
• Add documentation about similarity by @lukas-lightly in #188
• Allow creating and editing annotations within the frame details view by @LeonardoRosaa in #179
• Refactor _get_class_balancing_data() by @lukas-lightly in #191
• Improve video grid loading speed by @horatiualmasan in #183
• Remove DatasetTable.get_samples function by @masakljun in #184
• Fix video frame orientation by @michal-lightly in #192
• Display annotations within the video grid view and video details view by @LeonardoRosaa in #193
• Display annotations within the video frame grid view by @LeonardoRosaa in #189
• Minor docs improvements for lightlystudio by @IgorSusmelj in #186
• Rename Dataset.add_samples_from_path to Dataset.add_images_from_path by @horatiualmasan in #197
• Update docs and changelog with video by @michal-lightly in #195
• Hide annotations menu item for video and video frame sample types by @LeonardoRosaa in #196
• Add similarity API endpoint by @lukas-lightly in #190
• Fixed label disappearing after a few clicks by @LeonardoRosaa in #200
• Fix flakey test by @mihnea-necsulescu in #199
• Bump version to 0.4.4 by @michal-lightly in #202

Full Changelog: v0.4.3...v0.4.4

v0.4.3

[0.4.3] - 2025-11-13

Added

• Added class balancing with a uniform or the input distribution as target. These options can be set for the AnnotationClassBalancingStrategy.

Fixed

• Fixed installation issue with Python 3.13: Properly declare package compatibility in pyproject.toml.

What's Changed

• Add uniform class balancing by @lukas-lightly in #112
• Added endpoint to retrieve a video by @LeonardoRosaa in #116
• Add tag_depth=1 support to add_samples_from_path by @masakljun in #84
• Added endpoint to retrieve a frame by @LeonardoRosaa in #117
• Add videos from path by @horatiualmasan in #90
• Add caption update resolver by @JonasWurst in #120
• Reintroduce SampleFilter by @michal-lightly in #115
• Add "input" class balancing by @lukas-lightly in #125
• Add cloud storage installation info to docs by @masakljun in #126
• Bugfix: Use mock DB in tests by @michal-lightly in #129
• Add an example video script for development testing by @michal-lightly in #127
• Adding caption update endpoint by @JonasWurst in #128
• Rename samples backend route to images by @michal-lightly in #131
• Add dataset and caption filters to SampleFilter by @michal-lightly in #133
• Fix: Properly declare Python 3.13 compatibility. Bump version to 0.4.3. by @michal-lightly in #136

Full Changelog: v0.4.2...v0.4.3

v0.4.2

[0.4.2] - 2025-11-11

Added

• Added new selection strategy: AnnotationClassBalancingStrategy.
• Support for Python 3.13.
• Display captions within Sample Details.
• Added more detailed setup instructions to CONTRIBUTION.md
• Added a detailed section about cloud support to the docs.

Changed

• Changed the grid slider to define how many items will appear per row.
• Auto-scroll to the selected annotation in the sample details side panel.

Fixed

• Fixed issue when embedding plot wasn't updating after changing filters.
• Prevent duplicated annotation labels: Fixed an issue that occurred when adding samples from yolo using multiple splits.
• Added requests as an explicit dependency to prevent potential errors during embedding model download.
• Embedding generation RAM usage fixed by using np.ndarray.

What's Changed

• First step of refactoring the dataset resolver by @IgorSusmelj in #49
• Merge a feature branch with initial changes for multimodality support by @michal-lightly in #45
• Added the make command for downloading the example dataset by @LeonardoRosaa in #50
• Rename test helper create_sample to create_image by @michal-lightly in #48
• Rename variables in tests by @michal-lightly in #51
• Make thumbnail resizing more UX friendly by @LeonardoRosaa in #42
• Updated the panel slider style by @LeonardoRosaa in #53
• Introduce dataset hierarchy by @michal-lightly in #52
• Move dataset resolver export functions to a separate file by @michal-lightly in #54
• Add AnnotationClassBalancingStrategy by @lukas-lightly in #40
• Move the rest of dataset resolver functions to their own file by @michal-lightly in #56
• Auto-scroll to the selected annotation in the sample details side panel when selecting an annotation on a sample by @mihnea-necsulescu in #55
• Update the dataset_query imports by @JonasWurst in #57
• Rename datasets_resolver to dataset_resolver by @michal-lightly in #58
• Remove DatasetLoader class properly by @michal-lightly in #61
• Use pydantic mypy plugin for a stricter type check by @michal-lightly in #60
• Use np.ndarray for embed_images() by @lukas-lightly in #62
• Rename image_resolver to image_resolver_legacy by @michal-lightly in #64
• Remove API endpoint to create a dataset by @michal-lightly in #63
• Split image resolver by @michal-lightly in #67
• Create video and videoframe tables by @JonasWurst in #68
• Split image resolver tests by @michal-lightly in #72
• Introduce SampleType by @michal-lightly in #70
• feat: refactor Plot component to make it more controllable and scalable to extend by @ikondrat in #69
• Display captions within Sample Details by @LeonardoRosaa in #71
• Add video and video frame view with count by @JonasWurst in #74
• Check dataset type when adding samples by @michal-lightly in #78
• Added steps about how to set up the project for development by @LeonardoRosaa in #47
• Video create and get_all_by_dataset resolvers by @JonasWurst in #73
• Remove image_resolver.create by @michal-lightly in #79
• feat: update plot when filter is updated by @ikondrat in #80
• Return sample IDs from create_many by @michal-lightly in #81
• Rename SampleImage to ImageStub by @michal-lightly in #86
• Torch version handling by @lukas-lightly in #87
• Remove the image_resolver.update function by @michal-lightly in #88
• Remove dataset_id from ImageCreate by @michal-lightly in #89
• Add video resolver: count_by_id and filter_new_paths by @JonasWurst in #75
• Adding Basic Operator Parameters by @JonasWurst in #85
• Prevent adding annotation label duplicates by @JonasWurst in #92
• Remove dataset_id from ImageView by @michal-lightly in #93
• videoframe resolvers frame create get by @JonasWurst in #76
• Prepare the header for video support by @LeonardoRosaa in #91
• Introduce BaseOperator by @JonasWurst in #94
• Add explicit requests dependency by @michal-lightly in #98
• Use SampleTable.dataset_id instead of ImageTable.dataset_id in queries by @michal-lightly in #95
• Operator System: registry by @JonasWurst in #97
• Rename video table fields by @horatiualmasan in #99
• Add get_or_create_video_frame_child resolver function by @horatiualmasan in #101
• Remove dataset_id from ImageTable by @michal-lightly in #96
• Rename AnnotationBaseTable sample_id to parent_sample_id by @horatiualmasan in #102
• add to docs about using cloud storage and current support limitations by @masakljun in #100
• Rename Caption sample_id to parent_sample_id by @michal-lightly in #104
• Allow the usage of scikit-learn 1.7.2 by @lukas-lightly in #103
• Add a note about Linux and Firefox by @lukas-lightly in #109
• Split sample resolver by @michal-lightly in #107
• Added endpoint to retrieve a list of videos by @LeonardoRosaa in #105
• Use VIDEO_FRAME dataset type for video frame datasets by @michal-lightly in #113
• Added endpoint to retrieve a list of frames by @LeonardoRosaa in #106
• Update read_sample to use ImageView.ImageViewTag objects by @horatiualmasan in #110
• Remove dataset_id from AnnotationCreate by @horatiualmasan in #111
• Rename SampleFilter to ImageFilter by @michal-lightly in #114
• Add operator endpoints by @JonasWurst in #108
• Bump version to 0.4.2 by @JonasWurst in #121

New Contributors

• @lukas-lightly made their first contribution in #40
• @masakljun made their first contribution in #100

Full Changelog: v0.4.1...v0.4.2

v0.4.1

[0.4.1] - 2025-10-27

Added

• Added a footer with useful links and information about filtered and total annotations or samples.
• Improved class docstrings for the most important user-facing classes.
• Added Annotation tags section within the Annotation Details.
• Added undoable action for editing annotations on the sample details.
• Allowed users to remove Annotation tags from the Annotation Details.

Changed

• Updated button text to "View sample" in annotation details panel for better clarity.
• Pressing Escape while adding an annotation now cancels add-annotation mode.
• Improved the navbar to display button titles on hover and removed button text on small screens.
• Samples are now ordered by their filenames in the GUI.
• Introduce button to reset viewport changes for embedding plot.
• Improve UX for label picker when adding labels.

Removed

• Branding link from the Embedding Views status bar

What's Changed

• Add analytics to docs by @michal-lightly in #1
• Update Release Date by @JonasWurst in #2
• Make it easier to leave adding annotation mode on sample details by @ikondrat in #5
• Update link to sample details by @ikondrat in #3
• Embedding plot: Remove unnecessary stuff from status bar by @LeonardoRosaa in #4
• Add Status Bar with links to Docs and Support to the GUI by @MalteEbner in #14
• Add docstrings to DatasetQuery by @MalteEbner in #10
• Add docstrings to SampleField by @MalteEbner in #17
• Make navbar more user friendly by @LeonardoRosaa in #15
• Add docstrings to Selection class by @MalteEbner in #21
• Update export dialog by @horatiualmasan in #11
• Order samples in gui by filename update e2e tests by @mihnea-necsulescu in #20
• Use Dataset instead of DatasetLoader for e2e tests by @michal-lightly in #25
• Add issue templates by @horatiualmasan in #24
• Remove DatasetLoader by @michal-lightly in #26
• Introduce button to reset to default view in plot by @ikondrat in #27
• Regenerate schema by @michal-lightly in #30
• Remove unique from file_path_abs by @horatiualmasan in #31
• Add contributing rules by @horatiualmasan in #29
• Add tag section to annotation detail view by @LeonardoRosaa in #23
• Undo in GUI does not work for editing by @LeonardoRosaa in #32
• Rework FSC Create and Refine GUI by @horatiualmasan in #12
• Cache 2d-embeddings in the database. by @MalteEbner in #33
• Allow removing of annotation tags by @LeonardoRosaa in #39
• Improve UX for label picker when adding labels by @IgorSusmelj in #37
• Show number of filtered and total samples to footer by @IgorSusmelj in #38

Full Changelog: v0.4.0...v0.4.1

v0.4.0

Added

• Public LightlyStudio release

Full Changelog: https://github.com/lightly-ai/lightly-studio/commits/v0.4.0