โš™๏ธManage Datasets

Dataset Management and Version Control

In Unitlab Annotate, datasets are produced from projects based on the most recent annotations. Learn how to release datasets on the Release Dataset page.

๐Ÿ›ฐ๏ธRelease Datasets

On the Datasets page of Unitlab Annotate, you can manage your own datasets as well as explore and use public datasets shared by others. You can clone any public dataset and create your own version.

Learn how to clone a dataset and create your own version on the "Clone Datasets" page.

๐ŸŒ€Clone Datasets

In Unitlab Annotate, you can manage your datasets with the following actions:

Let's learn how to manage version, format and access control of datasets of Unitlab Annotate.

Version Control

Each new release creates a new version of the dataset. You can create an unlimited number of versions of your datasets from your projects as follows:

These releases create the versions of the dataset. To view your released dataset, go to the Datasets page in Unitlab Annotate and click on the dataset box. You can check the versions as shown in the screenshot below.

To view the annotation data of your dataset with the selected version, click on the "Data" section within your dataset page.

You can view each annotation based on the version you select. To see a visual annotation, click to open the data.

Format Control

In Unitlab Annotate, the following common export formats are currently available. Other formats will be available in the next version of Unitlab Annotate in the near future.

  • COCO

  • YOLOv5

  • YOLOv8

To download your dataset in the preferred format and version, visit your dataset page and click on the "Download" button located in the right corner as follows.

You can download your datasets using the following methods:

  • Download directly in a browser

  • Download via CLI

  • Download via Python SDK

If your datasets are large, we recommend using Unitlab's CLI/Python SDK tools.

Learn how to use the CLI/Python SDK tools of Unitlab Annotate by visiting the Unitlab CLI and Unitlab Python SDK pages.

Access Control

In Unitlab Annotate, datasets are divided into two access groups:

  • Private Datasets

  • Public Datasets.

Private datasets are always confidential and accessible only to the Owner (you) and the Managers of the workspace. Learn about member roles and role-based access control on the Role-Based Access Page.

โ›ณRole-Based Access

Public datasets are accessible to anyone who has registered and joined Unitlab Annotate. Public datasets can be switched from private to public, or initially released as public if you are using the Free Plan.

Please note that when you use Unitlab Annotate with the Free Plan, every dataset you release becomes publicly available on the "Dataset" page of Unitlab Annotate. Before releasing, you must agree to assign a license to your dataset. Learn how to safely release datasets and their licenses on the Release Datasets page.

๐Ÿ›ฐ๏ธRelease Datasets

To keep your released datasets private, consider upgrading your plan.

Discover Unitlab Annotate's Plans and Pricing

You can change the access settings or make your dataset public to anyone on the Dataset page of Unitlab Annotate by updating the "Settings" of your dataset as follows:"

To make your private dataset to public, you need to select a licence to your dataset. The public data you made is acknowledged under the name of your workspace and the license you select on the Dataset page of Unitlab Annotate.

Last updated