CHANGELOG.md

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[unreleased]

New features

Enhancements

Bug fixes

Q3 2024 Release 1.8.0

New features

• Add TabularValidator (#1498)
• Add Clean Transform for tabular data type (#1520)

Enhancements

• Set label name with parents to avoid duplicates for AstypeAnnotations (#1492)
• Pass Keyword Argument to TabularDataBase (#1522)
• Support hierarchical structure for ImageNet dataset format (#1528)
• Enable dtype argument when calling media.data (#1546)

Bug fixes

• Preserve end_frame information of a video when it is zero. (#1541)
• Changed the Datumaro format to ensure exported videos have relative paths and to prevent the same video from being overwritten. (#1547)

Q2 2024 Release 1.7.0

New features

• Support 'Video' media type in datumaro format (#1491)
• Add ann_types property for dataset (#1422, #1479)
• Add AnnotationType.rotated_bbox for oriented object detection (#1459)
• Add DOTA data format for oriented object detection task (#1475)
• Add AstypeAnnotations Transform (#1484)
• Enhance DatasetItem annotations for semantic segmentation model training use case (#1503)
• Add TabularValidator (#1498)
• Add Clean Transform for tabular data type (#1520)
• Add notebook for data handling of kaggle dataset (#1534)

Enhancements

• Fix ambiguous COCO format detector (#1442)
• Get target information for tabular dataset (#1471)
• Add ExtractedMask and update importers who can use it to use it (#1480)
• Improve PIL and COLOR_BGR context image decode performance (#1501)
• Improve get_area() of Polygon through Shoelace formula (#1507)
• Improve _Shape point converter (#1508)

Bug fixes

• Split the video directory into subsets to avoid overwriting (#1485)
• Doc update to replace --save-images is replaced with --save-media (#1514)

May 2024 Release 1.6.1

Enhancements

• Prevent AcLauncher for OpenVINO 2024.0 (#1450)

Bug fixes

• Modify lxml dependency constraint (#1460)
• Fix CLI error occurring when installed with default option only (#1444, #1454)
• Relax Pillow dependency constraint (#1436)
• Modify Numpy dependency constraint (#1435)
• Relax old pandas version constraint (#1467)

Apr. 2024 Release 1.6.0

New features

• Changed supported Python version range (>=3.9, <=3.11) (#1269)
• Support MMDetection COCO format (#1213)
• Develop JsonSectionPageMapper in Rust API (#1224)
• Add Filtering via User-Provided Python Functions (#1230, #1233)
• Remove supporting MacOS platform (#1235)
• Support Kaggle image data (KaggleImageCsvBase, KaggleImageTxtBase, KaggleImageMaskBase, KaggleVocBase, KaggleYoloBase) (#1240)
• Add __getitem__() for random accessing with O(1) time complexity (#1247)
• Add Data-aware Anchor Generator (#1251)
• Support bounding box import within Kaggle extractors and add KaggleCocoBase (#1273)

Enhancements

• Optimize Python import to make CLI entrypoint faster (#1182)
• Add ImageColorScale context manager (#1194)
• Enhance visualizer to toggle plot title visibility (#1228)
• Enhance Datumaro data format detect() to be memory-bounded and performant (#1229)
• Change RoIImage and MosaicImage to have np.uint8 dtype as default (#1245)
• Enable image backend and color channel format to be selectable (#1246)
• Boost up CityscapesBase and KaggleImageMaskBase by dropping np.unique (#1261)
• Enhance RISE algortihm for explainable AI (#1263)
• Enhance explore unit test to use real dataset from ImageNet (#1266)
• Fix each method of the comparator to be used separately (#1290)
• Bump ONNX version to 1.16.0 (#1376)
• Print the color channel format (RGB) for datum stats command (#1389)
• Add ignore_index argument to Mask.as_class_mask() and Mask.as_instance_mask() (#1409)

Bug fixes

• Fix wrong example of Datumaro dataset creation in document (#1195)
• Fix wrong command to install datumaro from github (#1202, #1207)
• Update document to correct wrong datum project import command and add filtering example to filter out items containing annotations. (#1210)
• Fix label compare of distance method (#1205)
• Fix Datumaro visualizer's import errors after introducing lazy import (#1220)
• Fix broken link to supported formats in readme (#1221)
• Fix Kinetics data format to have media data (#1223)
• Handling undefined labels at the annotation statistics (#1232)
• Add unit test for item rename (#1237)
• Fix a bug in the previous behavior when importing nested datasets in the project (#1243)
• Fix Kaggle importer when adding duplicated labels (#1244)
• Fix input tensor shape in model interpreter for OpenVINO 2023.3 (#1251)
• Add default value for target in prune cli (#1253)
• Remove deprecated MediaManager (#1262)
• Fix explore command without project (#1271)
• Fix enable COCO to import only bboxes (#1360)
• Fix resize transform for RleMask annotation
• (#1361)
• Fix import YOLO variants from extractor when urls is not specified (#1362)

Jan. 2024 Release 1.5.2

Enhancements

• Add memory bounded datumaro data format detect to release 1.5.1 (#1241)
• Bump version string to 1.5.2 (#1249)
• Remove Protobuf version limitation (<4) (#1248)

Nov. 2023 Release 1.5.1

Enhancements

• Enhance Datumaro data format stream importer performance (#1153)
• Change image default dtype from float32 to uint8 (#1175)
• Add comparison level-up doc (#1174)
• Add ImportError to catch GitPython import error (#1174)

Bug fixes

• Modify the draw function in the visualizer not to raise an error for unsupported annotation types. (#1180)
• Correct explore path in the related document. (#1176)
• Fix errata in the voc document. Color values in the labelmap.txt should be separated by commas, not colons. (#1162)
• Fix hyperlink errors in the document (#1159, #1161)
• Fix memory unbounded Arrow data format export/import (#1169)
• Update CVAT format doc to bypass warning (#1183)

15/09/2023 - Release 1.5.0

New features

• Add SAMAutomaticMaskGeneration transform (#1168)
• Add tabular data import/export (#1089)
• Support video annotation import/export (#1124)
• Add multiframework (PyTorch, Tensorflow) converter (#1125)
• Add SAM OVMS and Triton server Docker image builders (#1129)
• Add SAMBboxToInstanceMask transform (#1133, #1134)
• Add ConfigurableValidator (#1142)

Enhancements

• Enhance ClassificationValidator for multi-label classification datasets with label_groups (#1116)
• Replace Roboflow xml.etree with defusedxml (#1117)
• Define GroupType with IntEnum for, where 0 is EXCLUSIVE (#1116)
• Add Rust API to optimize COCOPageMapper performance (#1120)
• Support a dictionary input in addition to a single image input for the model launcher to support Segment Anything Model (#1133)
• Remove deprecates announced to be removed in 1.5.0 (#1140)
• Add multi-threading option to ModelTransform and SAMBboxToInstanceMask (#1145, #1149)

Bug fixes

• Coco exporter can export annotations even if there is no media, except for mask annotations which require media info. (#1147)(#1158)
• Fix bugs for Tile transform (#1123)
• Disable Roboflow Tfrecord format when Tensorflow is not installed (#1130)
• Raise VcsAlreadyExists error if vcs directory exists (#1138)

27/07/2023 - Release 1.4.1

Bug fixes

• Report errors for COCO (stream) and Datumaro importers (#1110)

21/07/2023 - Release 1.4.0

New features

• Add documentation and notebook example for Prune API (#1070)
• Changed supported Python version range (>=3.8, <=3.11) (#1083)
• Migrate OpenVINO v2023.0.0 (#1036)
• Add Roboflow data format support (COCO JSON, Pascal VOC XML, YOLOv5-PyTorch, YOLOv7-PyTorch, YOLOv8, YOLOv5 Oriented Bounding Boxes, Multiclass CSV, TFRecord, CreateML JSON) (#1044)
• Add MissingAnnotationDetection transform (#1049, #1063, #1064)
• Add OVMSLauncher (#1056)
• Add Prune API (#1058)
• Add TritonLauncher (#1059)
• Migrate DVC v3.0.0 (#1072)
• Stream dataset import/export (#1077, #1081, #1082, #1091, #1093, #1098, #1102)
• Support mask annotations for CVAT data format (#1078)

Enhancements

• Support list query for explorer (#1087)
• update contributing.md (#1094)
• Update 3rd-party.txt for release 1.4.0 (#1099)
• Give notice that the deprecation works will be done in datumaro==1.5.0 (#1085)
• Unify COCO, Datumaro, VOC, YOLO importer/exporter progress reporter descriptions (#1100)
• Enhance import performance for built-in plugins (#1031)
• Change default dtype of load_image() to np.uint8 (#1041)
• Add OTX ATSS detector model interpreter & refactor interfaces (#1047)
• Refactor Launcher and ModelInterpreter (#1055)
• Add CVAT data format document (#1060)
• Reduce peak memory usage when importing COCO and Datumaro formats (#1061)
• Enhance the error message for datum stats to be more user friendly (#1069)
• Refactor dataset.py to seperate DatasetStorage (#1073)

Bug fixes

• Create cache dir under only writable filesystem (#1088)
• Fix: Dataset infos() can be broken if a transform not redefining infos() is stacked on the top (#1101)
• Fix warnings in test_visualizer.py (#1039)
• Fix LabelMe data format (#1053)
• Prevent installing protobuf>=4 (#1054)
• Fix UnionMerge (#1086)

26/05/2023 - Release 1.3.2

Enhancements

• Let CocoBase continue even if an InvalidAnnotationError is raised (#1050)

Bug fixes

• Install dvc version to 2.x (#1048)
• Replace np.append() in Validator (#1050)

26/05/2023 - Release 1.3.1

Bug fixes

• Fix Cityscapes format mis-detection (#1029)

25/05/2023 - Release 1.3.0

New features

• Add CocoRoboflowImporter (#976, #1000)
• Add SynthiaSfImporter and SynthiaAlImporter (#987)
• Add intermediate skill docs for filter (#996)
• Add VocInstanceSegmentationImporter and VocInstanceSegmentationExporter (#997)
• Add Segment Anything data format support (#1005, #1009)
• Add Correct transformation (#1006)
• Implement ReindexAnnotations transform (#1008)
• Add notebook examples for importing/exporting detection and segmentation data (#1020, #1023)
• Update CLI from diff to compare, add TableComparator (#1012)

Enhancements

• Use autosummary for fully-automatic Python module docs generation (#973)
• Enrich stack trace for better user experience when importing (#992)
• Save and load hashkey for explorer (#981) (#1003)
• Add MOT and MOTS data format docs (#999)
• Improve RemoveAnnotations to remove specific annotations with ids (#1004)
• Add Jupyter notebook example of noisy label detection for detection tasks (#1011)

Bug fixes

• Fix Mapillary Vistas data format (#977)
• Fix bytes property returning None if function is given to data (#978)
• Fix Synthia-Rand data format (#987)
• Fix person_layout categories and action_classification attributes in imported Pascal-VOC dataset (#997)
• Drop a malformed transform from StackedTransform automatically (#1001)
• Fix Cityscapes to drop ImgsFine directory (#1023)

04/05/2023 - Release 1.2.1

Bug fixes

• Fix project level CVAT for images format import (#980)
• Fix an info message when using the convert CLI command with no args.input_format (#982)
• Fix media contents not returning bytes in arrow format (#986)

20/04/2023 - Release 1.2.0

New features

• Add Skill Up section to documentation (#920, #933, #935, #945, #949, #953, #959, #960, #967)
• Add LossDynamicsAnalyzer for noisy label detection (#928)
• Add Apache Arrow format support (#931, #948)
• Add sort transform (#931)

Enhancements

• Add multiprocessing to DatumaroBinaryBase (#897)
• Refactor merge code (#901, #906)
• Refactor download CLI commands (#909)
• Refactor CLI commands w/ and w/o project (#910, #952)
• Refactor Media to be initialized from explicit sources (#911 #921, #944)
• Refactor hl_ops.py (#912)
• Add tfds:uc_merced and tfds:eurosat download (#914)
• Migrate documentation framework to Sphinx (#917, #922, #947, #954, #958, #961, #962, #963, #964, #965, #969)
• Update merge tutorial for real life usecase (#930)
• Abbreviate "detect-format" to "detect" for prettifying (#951)

Bug fixes

• Add UserWarning if an invalid media_type comes to image statistics computation (#891)
• Fix negated is_encrypted (#907)
• Save extra images of PointCloud when exporting to datumaro format (#918)
• Fix log issue when importing celeba and align celeba dataset (#919)

28/03/2023 - Release 1.1.1

Bug fixes

• Fix to not export absolute media path in Datumaro and DatumaroBinary formats (#896)
• Change pypi_publish.yml to publish_sdist_to_pypi.yml (#895)

23/03/2023 - Release 1.1.0

New features

• Add with_subset_dirs decorator (Add ImagenetWithSubsetDirsImporter) (#816)
• Add CommonSemanticSegmentationWithSubsetDirsImporter (#826)
• Add DatumaroBinary format (#828, #829, #830, #831, #880, #883)
• Add Explorer CLI documentation (#838)
• Add version to dataset exported as datumaro format (#842)
• Add Ava action data format support (#847)
• Add Shift Analyzer (both covariate and label shifts) (#855)
• Add YOLO Loose format (#856)
• Add Ultralytics YOLO format (#859)

Enhancements

• Refactor Datumaro format code and test code (#824)
• Add publish to PyPI Github action (#867)
• Add --no-media-encryption option (#875)

Bug fixes

• Fix image filenames and anomaly mask appearance in MVTec exporter (#835)
• Fix CIFAR10 and 100 detect function (#836)
• Fix celeba and align_celeba detect function (#837)
• Choose the top priority detect format for all directory depths (#839)
• Fix MVTec format detect function (#843)
• Fix wrong __len__() of Subset when the item is removed (#854)
• Fix mask visualization bug (#860)
• Fix detect unit tests to test false negatives as well (#868)

24/02/2023 - Release v1.0.0

New features

• Add Data Explorer (#773)
• Add Ellipse annotation type (#807)
• Add MVTec anomaly data support (#810)

Enhancements

• Refactor existing tests (#803)
• Raise ImportError on importing malformed COCO directory (#812)
• Remove the duplicated and cyclical category context in documentation (#822)

Bug fixes

• Fix for importing CVAT image 1.1 data format exported to project level (#795)
• Fix a problem on setting log-level via CLI (#800)
• Fix code format with the latest black==23.1.0 (#802)
• Fix Explain command cannot find the model (#721) (#804)
• Fix a problem found on model remove CLI command (#805)

27/01/2023 - Release v0.5.0

New features

• Add Tile transformation (#790)
• Add Video keyframe extraction (#791)
• Add TileTransform documentation and Jupyter notebook example (#794)
• Add MergeTile transformation (#796)

Enhancements

• Improved mask_to_rle performance (#770)

Deprecated

• N/A

Removed

• N/A

Bug fixes

• Fix MacOS CI failures (#789)
• Fix auto-documentation for the data_format plugins (#793)

Security

• Add security.md file for the SDL (#798)

06/12/2022 - Release v0.4.0.1

New features

• Support for exclusive of labels with LabelGroup (#742)
• Jupyter samples
  • Introducing how to merge datasets (#738)
  • Introducing how to visualize dataset (#747)
  • Introducing how to filter dataset (#748)
  • Introducing how to transform dataset (#759)
• Visualization Python API
  • Bbox feature (#744)
  • Label, Points, Polygon, PolyLine, and Caption visualization features (#746)
  • Mask, SuperResolution, Depth visualization features (#747)
• Documentation for Python API (#753)
  • dataset handler, visualizer, filter descriptions (#761)
• __repr__ for Dataset (#750)
• Support for exporting as CVAT video format (#757)
• CodeCov coverage reporting feature to CI/CD (#756)
• Jupyter notebook example rendering to documentation (#758)
• An interface to manipulate 'infos' to store the dataset meta-info (#767)
• 'bbox' annotation when importing a COCO dataset (#772)

Enhancements

• Wrap title text according to its plot width (#769)
• Get list of subsets and support only Image media type in visualizer (#768)

Deprecated

• N/A

Removed

• N/A

Bug fixes

• Correcting static type checking (#743)
• Fixing a VOC dataset export when a label contains 'space' (#771)

Security

• N/A

06/09/2022 - Release v0.3.1

New features

• Support for custom media types, new PointCloud media type, DatasetItem.media and .media_as(type) members (#539)
• [API] A way to request dataset and extractor media type with media_type (#539)
• BraTS format (import-only) (.npy and .nii.gz), new MultiframeImage media type (#628)
• Common Semantic Segmentation dataset format (import-only) (#685)
• An option to disable data/ prefix inclusion in YOLO export (#689)
• New command describe-downloads to print information about downloadable datasets (#678)
• Detection for Cityscapes format (#680)
• Maximum recursion --depth parameter for detect-dataset CLI command (#680)
• An option to save a single subset in the download command (#697)
• Common Super Resolution dataset format (import-only) (#700)
• Kinetics 400/600/700 dataset format (import-only) (#706)
• NYU Depth Dataset V2 format (import-only) (#712)

Enhancements

• env.detect_dataset() now returns a list of detected formats at all recursion levels instead of just the lowest one (#680)
• Open Images: allowed to store annotations file in root path as well (#680)
• Improved parsing error messages in COCO, VOC and YOLO formats (#684, #686, #687)
• YOLO format now supports almost any subset names, except backup, names and classes (instead of just train and valid). The reserved names now raise an error on exporting. (#688)

Deprecated

• --save-images is replaced with --save-media in CLI and converter API (#539)
• [API] image, point_cloud and related_images of DatasetItem are replaced with media and media_as(type) members and c-tor parameters (#539)

Removed

• N/A

Bug fixes

• Detection for LFW format (#680)
• Adding depth value of image when dataset is exported in VOC format (#726)
• Adding to handle the numerical labels in task chains properly (#726)
• Fixing the issue that annotations inside another annotation (polygon) are duplicated during import for VOC format (#726)

Security

• N/A

21/02/2022 - Release v0.3

New features

• Ability to import a video as frames with the video_frames format and to split a video into frames with the datum util split_video command (#555)
• --subset parameter in the image_dir format (#555)
• MediaManager API to control loaded media resources at runtime (#555)
• Command to detect the format of a dataset (#576)
• More comfortable access to library API via import datumaro (#630)
• CLI command-like free functions (export, transform, ...) (#630)
• Reading specific annotation files for train dataset in Cityscapes (#632)
• Random sampling transforms (random_sampler, label_random_sampler) to create smaller datasets from bigger ones (#636, #640)
• API to report dataset import and export progress; API to report dataset import and export errors and take action (skip, fail) (supported in COCO, VOC and YOLO formats) (#650)
• Support for downloading the ImageNetV2 and COCO datasets (#653, #659)
• A way for formats to signal that they don't support detection (#665)
• Removal transforms to remove items/annoations/attributes from dataset (remove_items, remove_annotations, remove_attributes) (#670)

Enhancements

• Allowed direct file paths in datum import. Such sources are imported like when the rpath parameter is specified, however, only the selected path is copied into the project (#555)
• Improved stats performance, added new filtering parameters, image stats (unique, repeated) moved to the dataset section, removed mean and std from the dataset section (#621)
• Allowed Image creation from just size info (#634)
• Added image search in VOC XML-based subformats (#634)
• Added image path equality checks in simple merge, when applicable (#634)
• Supported saving box attributes when downloading the TFDS version of VOC (#668)
• Switched to a pyproject.toml-based build (#671)

Deprecated

• TBD

Removed

• Official support of Python 3.6 (due to it's EOL) (#617)
• Backward compatibility annotation symbols in components.extractor (#630)

Bug fixes

• Prohibited calling add, import and export commands without a project (#555)
• Calling make_dataset on empty project tree now produces the error properly (#555)
• Saving (overwriting) a dataset in a project when rpath is used (#613)
• Output image extension preserving in the Resize transform (#606)
• Memory overuse in the Resize transform (#607)
• Invalid image pixels produced by the Resize transform (#618)
• Numeric warnings that sometimes occurred in stats command (e.g. #607) (#621)
• Added missing item attribute merging in simple merge (#634)
• Inability to disambiguate VOC from LabelMe in some cases (#658)

Security

• TBD

28/01/2022 - Release v0.2.3

New features

• Command to download public datasets (#582)
• Extension autodetection in ByteImage (#595)
• MPII Human Pose Dataset (import-only) (.mat and .json) (#584)
• MARS format (import-only) (#585)

Enhancements

• The pycocotools dependency lower bound is raised to 2.0.4. (#449)
• smooth_line from datumaro.util.annotation_util - the function is renamed to approximate_line and has updated interface (#592)

Deprecated

• Python 3.6 support

Removed

• TBD

Bug fixes

• Fails in multimerge when lines are not approximated and when there are no label categories (#592)
• Cannot convert LabelMe dataset, that has no subsets (#600)

Security

• TBD

24/12/2021 - Release v0.2.2

New features

• Video reading API (#521)
• Python API documentation (#526)
• Mapillary Vistas dataset format (Import-only) (#537)
• Datumaro can now be installed on Windows on Python 3.9 (#547)
• Import for SYNTHIA dataset format (#532)
• Support of score attribute in KITTI detetion (#571)
• Support for Accuracy Checker dataset meta files in formats (#553, #569, #575)
• Import for VoTT dataset format (#573)
• Image resizing transform (#581)

Enhancements

• The following formats can now be detected unambiguously: ade20k2017, ade20k2020, camvid, coco, cvat, datumaro, icdar_text_localization, icdar_text_segmentation, icdar_word_recognition, imagenet_txt, kitti_raw, label_me, lfw, mot_seq, open_images, vgg_face2, voc, widerface, yolo (#531, #536, #550, #557, #558)
• Allowed Pytest-native tests (#563)
• Allowed export options in the datum merge command (#545)

Deprecated

• Using Image, ByteImage from datumaro.util.image - these classes are moved to datumaro.components.media (#538)

Removed

• Equality comparison support between datumaro.components.media.Image and numpy.ndarray (#568)

Bug fixes

• Bug #560: import issue with MOT dataset when using seqinfo.ini file (#564)
• Empty lines in VOC subset lists are not ignored (#587)

Security

• TBD

16/11/2021 - Release v0.2.1

New features

• Import for CelebA dataset format. (#484)

Enhancements

• File people.txt became optional in LFW (#509)
• File image_ids_and_rotation.csv became optional Open Images (#509)
• Allowed underscores (_) in subset names in COCO (#509)
• Allowed annotation files with arbitrary names in COCO (#509)
• The icdar_text_localization format is no longer detected in every directory (#531)
• Updated pycocotools version to 2.0.2 (#534)

Deprecated

• TBD

Removed

• TBD

Bug fixes

• Unhandled exception when a file is specified as the source for a COCO or MOTS dataset (#530)
• Exporting dataset without color attribute into the icdar_text_segmentation format (#556)

Security

• TBD

14/10/2021 - Release v0.2

New features

• A new installation target: pip install datumaro[default], which should be used by default. The simple datumaro is supposed for library users. (#238)
• Dataset and project versioning capabilities (Git-like) (#238)
• "dataset revpath" concept in CLI, allowing to pass a dataset path with the dataset format in diff, merge, explain and info CLI commands (#238)
• import, remove, commit, checkout, log, status, info CLI commands (#238)
• Coco*Extractor classes now have an option to preserve label IDs from the original annotation file (#453)
• patch CLI command to patch datasets (#401)
• ProjectLabels transform to change dataset labels for merging etc. (#401, #478)
• Support for custom labels in the KITTI detection format (#481)
• Type annotations and docs for Annotation classes (#493)
• Options to control label loading behavior in imagenet_txt import (#434, #489)

Enhancements

• A project can contain and manage multiple datasets instead of a single one. CLI operations can be applied to the whole project, or to separate datasets. Datasets are modified inplace, by default (#328)
• CLI help for builtin plugins doesn't require project (#328)
• Annotation-related classes were moved into a new module, datumaro.components.annotation (#439)
• Rollback utilities replaced with Scope utilities (#444)
• The Project class from datumaro.components is changed completely (#238)
• diff and ediff are joined into a single diff CLI command (#238)
• Projects use new file layout, incompatible with old projects. An old project can be updated with datum project migrate (#238)
• Inheriting CliPlugin is not required in plugin classes (#238)
• Importers do not create Projects anymore and just return a list of extractor configurations (#238)

Deprecated

• TBD

Removed

• import, project merge CLI commands (#238)
• Support for project hierarchies. A project cannot be a source anymore (#238)
• Project cannot have independent internal dataset anymore. All the project data must be stored in the project data sources (#238)
• datumaro_project format (#238)
• Unused path field of DatasetItem (#455)

Bug fixes

• Deprecation warning in open_images_format.py (#440)
• lazy_image returning unrelated data sometimes (#409)
• Invalid call to pycocotools.mask.iou (#450)
• Importing of Open Images datasets without image data (#463)
• Return value type in Dataset.is_modified (#401)
• Remapping of secondary categories in RemapLabels (#401)
• VOC dataset patching for classification and segmentation tasks (#478)
• Exported mask label ids in KITTI segmentation (#481)
• Missing label for Points read in the LFW format (#494)

Security

• TBD

24/08/2021 - Release v0.1.11

New features

• The Open Images format now supports bounding box and segmentation mask annotations (#352, #388).
• Bounding boxes values decrement transform (#366)
• Improved error reporting in Dataset (#386)
• Support ADE20K format (import only) (#400)
• Documentation website at https://openvinotoolkit.github.io/datumaro (#420)

Enhancements

• Datumaro no longer depends on scikit-image (#379)
• Dataset remembers export options on saving / exporting for the first time (#386)

Deprecated

• TBD

Removed

• TBD

Bug fixes

• Application of remap_labels to dataset categories of different length (#314)
• Patching of datasets in formats (#348)
• Improved Cityscapes export performance (#367)
• Incorrect format of *_labelIds.png in Cityscapes export (#325, #342)
• Item id in ImageNet format (#371)
• Double quotes for ICDAR Word Recognition (#375)
• Wrong display of builtin formats in CLI (#332)
• Non utf-8 encoding of annotation files in Market-1501 export (#392)
• Import of ICDAR, PASCAL VOC and VGGFace2 images from subdirectories on WIndows (#392)
• Saving of images with Unicode paths on Windows (#392)
• Calling ProjectDataset.transform() with a string argument (#402)
• Attributes casting for CVAT format (#403)
• Loading of custom project plugins (#404)
• Reading, writing anno file and saving name of the subset for test subset (#447)

Security

• Fixed unsafe unpickling in CIFAR import (#362)

14/07/2021 - Release v0.1.10

New features

• Support for import/export zip archives with images (#273)
• Subformat importers for VOC and COCO (#281)
• Support for KITTI dataset segmentation and detection format (#282)
• Updated YOLO format user manual (#295)
• ItemTransform class, which describes item-wise dataset Transforms (#297)
• keep-empty export parameter in VOC format (#297)
• A base class for dataset validation plugins (#299)
• Partial support for the Open Images format; only images and image-level labels can be read/written (#291, #315).
• Support for Supervisely Point Cloud dataset format (#245, #353)
• Support for KITTI Raw / Velodyne Points dataset format (#245)
• Support for CIFAR-100 and documentation for CIFAR-10/100 (#301)

Enhancements

• Tensorflow AVX check is made optional in API and disabled by default (#305)
• Extensions for images in ImageNet_txt are now mandatory (#302)
• Several dependencies now have lower bounds (#308)

Deprecated

• TBD

Removed

• TBD

Bug fixes

• Incorrect image layout on saving and a problem with ecoding on loading (#284)
• An error when XPath filter is applied to the dataset or its subset (#259)
• Tracking of Dataset changes done by transforms (#297)
• Improved CLI startup time in several cases (#306)

Security

• Known issue: loading CIFAR can result in arbitrary code execution (#327)

03/06/2021 - Release v0.1.9

New features

• Support for escaping in attribute values in LabelMe format (#49)
• Support for Segmentation Splitting (#223)
• Support for CIFAR-10/100 dataset format (#225, #243)
• Support for COCO panoptic and stuff format (#210)
• Documentation file and integration tests for Pascal VOC format (#228)
• Support for MNIST and MNIST in CSV dataset formats (#234)
• Documentation file for COCO format (#241)
• Documentation file and integration tests for YOLO format (#246)
• Support for Cityscapes dataset format (#249)
• Support for Validator configurable threshold (#250)

Enhancements

• LabelMe format saves dataset items with their relative paths by subsets without changing names (#200)
• Allowed arbitrary subset count and names in classification and detection splitters (#207)
• Annotation-less dataset elements are now participate in subset splitting (#211)
• Classification task in LFW dataset format (#222)
• Testing is now performed with pytest instead of unittest (#248)

Deprecated

• TBD

Removed

• TBD

Bug fixes

• Added support for auto-merging (joining) of datasets with no labels and having labels (#200)
• Allowed explicit label removal in remap_labels transform (#203)
• Image extension in CVAT format export (#214)
• Added a label "face" for bounding boxes in Wider Face (#215)
• Allowed adding "difficult", "truncated", "occluded" attributes when converting to Pascal VOC if these attributes are not present (#216)
• Empty lines in YOLO annotations are ignored (#221)
• Export in VOC format when no image info is available (#239)
• Fixed saving attribute in WiderFace extractor (#251)

Security

• TBD

31/03/2021 - Release v0.1.8

New features

• TBD

Enhancements

• Added an option to allow undeclared annotation attributes in CVAT format export (#192)
• COCO exports images in separate dirs by subsets. Added an option to control this (#195)

Deprecated

• TBD

Removed

• TBD

Bug fixes

• Instance masks of background class no more introduce an instance (#188)
• Added support for label attributes in Datumaro format (#192)

Security

• TBD

24/03/2021 - Release v0.1.7

New features

• OpenVINO plugin examples (#159)
• Dataset validation for classification and detection datasets (#160)
• Arbitrary image extensions in formats (import and export) (#166)
• Ability to set a custom subset name for an imported dataset (#166)
• CLI support for NDR(#178)

Enhancements

• Common ICDAR format is split into 3 sub-formats (#174)

Deprecated

• TBD

Removed

• TBD

Bug fixes

• The ability to work with file names containing Cyrillic and spaces (#148)
• Image reading and saving in ICDAR formats (#174)
• Unnecessary image loading on dataset saving (#176)
• Allowed spaces in ICDAR captions (#182)
• Saving of masks in VOC when masks are not requested (#184)

Security

• TBD

03/02/2021 - Release v0.1.6.1 (hotfix)

New features

• TBD

Enhancements

• TBD

Deprecated

• TBD

Removed

• TBD

Bug fixes

• Images with no annotations are exported again in VOC formats (#123)
• Inference result for only one output layer in OpenVINO launcher (#125)

Security

• TBD

02/26/2021 - Release v0.1.6

New features

• Icdar13/15 dataset format (#96)
• Laziness, source caching, tracking of changes and partial updating for Dataset (#102)
• Market-1501 dataset format (#108)
• LFW dataset format (#110)
• Support of polygons' and masks' confusion matrices and mismathing classes in diff command (#117)
• Add near duplicate image removal plugin (#113)
• Sampler Plugin that analyzes inference result from the given dataset and selects samples for annotation(#115)

Enhancements

• OpenVINO model launcher is updated for OpenVINO r2021.1 (#100)

Deprecated

• TBD

Removed

• TBD

Bug fixes

• High memory consumption and low performance of mask import/export, #53 (#101)
• Masks, covered by class 0 (background), should be exported with holes inside (#104)
• diff command invocation problem with missing class methods (#117)

Security

• TBD

01/23/2021 - Release v0.1.5

New features

• WiderFace dataset format (#65, #90)
• Function to transform annotations to labels (#66)
• Dataset splits for classification, detection and re-id tasks (#68, #81)
• VGGFace2 dataset format (#69, #82)
• Unique image count statistic (#87)
• Installation with pip by name datumaro

Enhancements

• Dataset class extended with new operations: save, load, export, import_from, detect, run_model (#71)
• Allowed importing Extractor-only defined formats (in Project.import_from, dataset.import_from and CLI/project import) (#71)
• datum project ... commands replaced with datum ... commands (#84)
• Supported more image formats in ImageNet extractors (#85)
• Allowed adding Importer-defined formats as project sources (source add) (#86)
• Added max search depth in ImageDir format and importers (#86)

Deprecated

• datum project ... CLI context (#84)

Removed

• TBD

Bug fixes

• Allow plugins inherited from Extractor (instead of only SourceExtractor) (#70)
• Windows installation with pip for pycocotools (#73)
• YOLO extractor path matching on Windows (#73)
• Fixed inplace file copying when saving images (#76)
• Fixed labelmap parameter type checking in VOC converter (#76)
• Fixed model copying on addition in CLI (#94)

Security

• TBD

12/10/2020 - Release v0.1.4

New features

• CamVid dataset format (#57)
• Ability to install opencv-python-headless dependency with DATUMARO_HEADLESS=1 environment variable instead of opencv-python (#62)

Enhancements

• Allow empty supercategory in COCO (#54)
• Allow Pascal VOC to search in subdirectories (#50)

Deprecated

• TBD

Removed

• TBD

Bug fixes

• TBD

Security

• TBD

10/28/2020 - Release v0.1.3

New features

• ImageNet and ImageNetTxt dataset formats (#41)

Enhancements

• TBD

Deprecated

• TBD

Removed

• TBD

Bug fixes

• Default label-map parameter value for VOC converter (#34)
• Randomness of random split transform (#38)
• Transform.subsets() method (#38)
• Supported unknown image formats in TF Detection API converter (#40)
• Supported empty attribute values in CVAT extractor (#45)

Security

• TBD

10/05/2020 - Release v0.1.2

New features

• ByteImage class to represent encoded images in memory and avoid recoding on save (#27)

Enhancements

• Implementation of format plugins simplified (#22)
• default is now a default subset name, instead of None. The values are interchangeable. (#22)
• Improved performance of transforms (#22)

Deprecated

• TBD

Removed

• image/depth value from VOC export (#27)

Bug fixes

• Zero division errors in dataset statistics (#31)

Security

• TBD

09/24/2020 - Release v0.1.1

New features

• reindex option in COCO and CVAT converters (#18)
• Support for relative paths in LabelMe format (#19)
• MOTS png mask format support (https://github.com/openvinotoolkit/datumaro/21)

Enhancements

• TBD

Deprecated

• TBD

Removed

• TBD

Bug fixes

• TBD

Security

• TBD

09/10/2020 - Release v0.1.0

New features

• Initial release

Template

## [Unreleased]
### New features
- TBD

### Enhancements
- TBD

### Deprecated
- TBD

### Removed
- TBD

### Bug fixes
- TBD

### Security
- TBD