# Dataset Management Framework (Datumaro)
[](https://github.com/openvinotoolkit/datumaro/actions/workflows/health_check.yml)
[](https://codecov.io/gh/openvinotoolkit/datumaro)
A framework and CLI tool to build, transform, and analyze datasets.
<!--lint disable fenced-code-flag-->
```
VOC dataset ---> Annotation tool
+ /
COCO dataset -----> Datumaro ---> dataset ------> Model training
+ \
CVAT annotations ---> Publication, statistics etc.
```
<!--lint enable fenced-code-flag-->
- [Getting started](https://openvinotoolkit.github.io/datumaro/docs/getting_started)
- [Features](#features)
- [User manual](https://openvinotoolkit.github.io/datumaro/docs/user-manual)
- [Developer manual](https://openvinotoolkit.github.io/datumaro/api)
- [Contributing](#contributing)
## Features
[(Back to top)](#dataset-management-framework-datumaro)
- Dataset reading, writing, conversion in any direction.
- [CIFAR-10/100](https://www.cs.toronto.edu/~kriz/cifar.html) (`classification`)
- [Cityscapes](https://www.cityscapes-dataset.com/)
- [COCO](http://cocodataset.org/#format-data) (`image_info`, `instances`, `person_keypoints`,
`captions`, `labels`, `panoptic`, `stuff`)
- [CVAT](https://openvinotoolkit.github.io/cvat/docs/manual/advanced/xml_format)
- [ImageNet](http://image-net.org/)
- [Kitti](http://www.cvlibs.net/datasets/kitti/index.php) (`segmentation`, `detection`,
`3D raw` / `velodyne points`)
- [LabelMe](http://labelme.csail.mit.edu/Release3.0)
- [LFW](http://vis-www.cs.umass.edu/lfw/) (`classification`, `person re-identification`,
`landmarks`)
- [MNIST](http://yann.lecun.com/exdb/mnist/) (`classification`)
- [Open Images](https://storage.googleapis.com/openimages/web/download.html)
- [PASCAL VOC](http://host.robots.ox.ac.uk/pascal/VOC/voc2012/htmldoc/index.html)
(`classification`, `detection`, `segmentation`, `action_classification`, `person_layout`)
- [TF Detection API](https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/using_your_own_dataset.md)
(`bboxes`, `masks`)
- [YOLO](https://github.com/AlexeyAB/darknet#how-to-train-pascal-voc-data) (`bboxes`)
Other formats and documentation for them can be found [here](https://openvinotoolkit.github.io/datumaro/docs/user-manual/supported_formats).
- Dataset building
- Merging multiple datasets into one
- Dataset filtering by a custom criteria:
- remove polygons of a certain class
- remove images without annotations of a specific class
- remove `occluded` annotations from images
- keep only vertically-oriented images
- remove small area bounding boxes from annotations
- Annotation conversions, for instance:
- polygons to instance masks and vice-versa
- apply a custom colormap for mask annotations
- rename or remove dataset labels
- Splitting a dataset into multiple subsets like `train`, `val`, and `test`:
- random split
- task-specific splits based on annotations,
which keep initial label and attribute distributions
- for classification task, based on labels
- for detection task, based on bboxes
- for re-identification task, based on labels,
avoiding having same IDs in training and test splits
- Sampling a dataset
- analyzes inference result from the given dataset
and selects the ‘best’ and the ‘least amount of’ samples for annotation.
- Select the sample that best suits model training.
- sampling with Entropy based algorithm
- Dataset quality checking
- Simple checking for errors
- Comparison with model inference
- Merging and comparison of multiple datasets
- Annotation validation based on the task type(classification, etc)
- Dataset comparison
- Dataset statistics (image mean and std, annotation statistics)
- Model integration
- Inference (OpenVINO, Caffe, PyTorch, TensorFlow, MxNet, etc.)
- Explainable AI ([RISE algorithm](https://arxiv.org/abs/1806.07421))
- RISE for classification
- RISE for object detection
> Check
[the design document](https://openvinotoolkit.github.io/datumaro/docs/design)
for a full list of features.
> Check
[the user manual](https://openvinotoolkit.github.io/datumaro/docs/user-manual)
for usage instructions.
## Contributing
[(Back to top)](#dataset-management-framework-datumaro)
Feel free to
[open an Issue](https://github.com/openvinotoolkit/datumaro/issues/new), if you
think something needs to be changed. You are welcome to participate in
development, instructions are available in our
[contribution guide](https://openvinotoolkit.github.io/datumaro/docs/contributing).
## Telemetry data collection note
The [OpenVINO™ telemetry library](https://github.com/openvinotoolkit/telemetry/)
is used to collect basic information about Datumaro usage.
To enable/disable telemetry data collection please see the
[guide](https://openvinotoolkit.github.io/datumaro/docs/user-manual/how_to_control_tm_data_collection/).
Raw data
{
"_id": null,
"home_page": "https://github.com/openvinotoolkit/datumaro",
"name": "datumaro-light",
"maintainer": "",
"docs_url": null,
"requires_python": ">=3.7",
"maintainer_email": "",
"keywords": "",
"author": "Intel",
"author_email": "emily.chun@intel.com",
"download_url": "https://files.pythonhosted.org/packages/93/ff/48eba8879c3e1eca1edf53b13a08aef18fe6f89521c1c549011b08954d3d/datumaro-light-1.2.1.dev1684947285.tar.gz",
"platform": null,
"description": "# Dataset Management Framework (Datumaro)\n\n[](https://github.com/openvinotoolkit/datumaro/actions/workflows/health_check.yml)\n[](https://codecov.io/gh/openvinotoolkit/datumaro)\n\nA framework and CLI tool to build, transform, and analyze datasets.\n\n<!--lint disable fenced-code-flag-->\n```\nVOC dataset ---> Annotation tool\n + /\nCOCO dataset -----> Datumaro ---> dataset ------> Model training\n + \\\nCVAT annotations ---> Publication, statistics etc.\n```\n<!--lint enable fenced-code-flag-->\n\n- [Getting started](https://openvinotoolkit.github.io/datumaro/docs/getting_started)\n- [Features](#features)\n- [User manual](https://openvinotoolkit.github.io/datumaro/docs/user-manual)\n- [Developer manual](https://openvinotoolkit.github.io/datumaro/api)\n- [Contributing](#contributing)\n\n## Features\n\n[(Back to top)](#dataset-management-framework-datumaro)\n\n- Dataset reading, writing, conversion in any direction.\n - [CIFAR-10/100](https://www.cs.toronto.edu/~kriz/cifar.html) (`classification`)\n - [Cityscapes](https://www.cityscapes-dataset.com/)\n - [COCO](http://cocodataset.org/#format-data) (`image_info`, `instances`, `person_keypoints`,\n `captions`, `labels`, `panoptic`, `stuff`)\n - [CVAT](https://openvinotoolkit.github.io/cvat/docs/manual/advanced/xml_format)\n - [ImageNet](http://image-net.org/)\n - [Kitti](http://www.cvlibs.net/datasets/kitti/index.php) (`segmentation`, `detection`,\n `3D raw` / `velodyne points`)\n - [LabelMe](http://labelme.csail.mit.edu/Release3.0)\n - [LFW](http://vis-www.cs.umass.edu/lfw/) (`classification`, `person re-identification`,\n `landmarks`)\n - [MNIST](http://yann.lecun.com/exdb/mnist/) (`classification`)\n - [Open Images](https://storage.googleapis.com/openimages/web/download.html)\n - [PASCAL VOC](http://host.robots.ox.ac.uk/pascal/VOC/voc2012/htmldoc/index.html)\n (`classification`, `detection`, `segmentation`, `action_classification`, `person_layout`)\n - [TF Detection API](https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/using_your_own_dataset.md)\n (`bboxes`, `masks`)\n - [YOLO](https://github.com/AlexeyAB/darknet#how-to-train-pascal-voc-data) (`bboxes`)\n\n Other formats and documentation for them can be found [here](https://openvinotoolkit.github.io/datumaro/docs/user-manual/supported_formats).\n- Dataset building\n - Merging multiple datasets into one\n - Dataset filtering by a custom criteria:\n - remove polygons of a certain class\n - remove images without annotations of a specific class\n - remove `occluded` annotations from images\n - keep only vertically-oriented images\n - remove small area bounding boxes from annotations\n - Annotation conversions, for instance:\n - polygons to instance masks and vice-versa\n - apply a custom colormap for mask annotations\n - rename or remove dataset labels\n - Splitting a dataset into multiple subsets like `train`, `val`, and `test`:\n - random split\n - task-specific splits based on annotations,\n which keep initial label and attribute distributions\n - for classification task, based on labels\n - for detection task, based on bboxes\n - for re-identification task, based on labels,\n avoiding having same IDs in training and test splits\n - Sampling a dataset\n - analyzes inference result from the given dataset\n and selects the \u2018best\u2019 and the \u2018least amount of\u2019 samples for annotation.\n - Select the sample that best suits model training.\n - sampling with Entropy based algorithm\n- Dataset quality checking\n - Simple checking for errors\n - Comparison with model inference\n - Merging and comparison of multiple datasets\n - Annotation validation based on the task type(classification, etc)\n- Dataset comparison\n- Dataset statistics (image mean and std, annotation statistics)\n- Model integration\n - Inference (OpenVINO, Caffe, PyTorch, TensorFlow, MxNet, etc.)\n - Explainable AI ([RISE algorithm](https://arxiv.org/abs/1806.07421))\n - RISE for classification\n - RISE for object detection\n\n> Check\n [the design document](https://openvinotoolkit.github.io/datumaro/docs/design)\n for a full list of features.\n> Check\n [the user manual](https://openvinotoolkit.github.io/datumaro/docs/user-manual)\n for usage instructions.\n\n## Contributing\n\n[(Back to top)](#dataset-management-framework-datumaro)\n\nFeel free to\n[open an Issue](https://github.com/openvinotoolkit/datumaro/issues/new), if you\nthink something needs to be changed. You are welcome to participate in\ndevelopment, instructions are available in our\n[contribution guide](https://openvinotoolkit.github.io/datumaro/docs/contributing).\n\n## Telemetry data collection note\n\nThe [OpenVINO\u2122 telemetry library](https://github.com/openvinotoolkit/telemetry/)\nis used to collect basic information about Datumaro usage.\n\nTo enable/disable telemetry data collection please see the\n[guide](https://openvinotoolkit.github.io/datumaro/docs/user-manual/how_to_control_tm_data_collection/).\n",
"bugtrack_url": null,
"license": "",
"summary": "Dataset Management Framework (Datumaro)",
"version": "1.2.1.dev1684947285",
"project_urls": {
"Homepage": "https://github.com/openvinotoolkit/datumaro"
},
"split_keywords": [],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "93ff48eba8879c3e1eca1edf53b13a08aef18fe6f89521c1c549011b08954d3d",
"md5": "a9d73ad28ef2e9aafa6742157ce049d2",
"sha256": "8f2f21152c1fe18ecf8cdc5f1da2fdcdc831dfac13f1c1ce8fac4d35abb25bf7"
},
"downloads": -1,
"filename": "datumaro-light-1.2.1.dev1684947285.tar.gz",
"has_sig": false,
"md5_digest": "a9d73ad28ef2e9aafa6742157ce049d2",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.7",
"size": 237023,
"upload_time": "2023-05-24T16:55:10",
"upload_time_iso_8601": "2023-05-24T16:55:10.892789Z",
"url": "https://files.pythonhosted.org/packages/93/ff/48eba8879c3e1eca1edf53b13a08aef18fe6f89521c1c549011b08954d3d/datumaro-light-1.2.1.dev1684947285.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2023-05-24 16:55:10",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "openvinotoolkit",
"github_project": "datumaro",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"requirements": [],
"tox": true,
"lcname": "datumaro-light"
}
JSON
