Release Datasets
Release datasets from your projects
Last updated
Was this helpful?
Release datasets from your projects
Last updated
Was this helpful?
In Unitlab Annotate, you can create unlimited datasets from your active projects. You can easily release annotations as datasets from projects. To release a dataset, you should have an active project with annotations.
To release a dataset in Unitlab Annotate, the following steps are required from your side.
Make sure you have projects or create a project and start annotating data.
Ensure the project has progress and annotations.
Navigate to the "Release Dataset" page within your project in Unitlab Annotate.
Select the preferred format (COCO, YOLOv5, or YOLOv8) for the dataset.
Click on 'Release' to create a new dataset version based on the latest annotations.
Manage Datasets on the Datasets page of Unitlab Annotate.
In Unitlab Annotate, datasets are produced from projects based on the most recent annotations. Please make sure that your projects contain annotated data.
In our demo example, we have experimentally created a project named "Vegetable and Fruit Segmentation" on the Setup a Project page and annotated its data on the Annotation page.
The progress of the project is 100% annotated and 100% reviewed. Let's release a dataset based on this progress. You can release datasets as soon as your projects contain annotated data.
To release a dataset, click on the project box and navigate to the "Release dataset" page of the project as shown in the screenshot below.
To make a new release, click the button in the right corner labeled "Release Dataset."
When releasing a dataset, you have the option to select either just the export format or both the export format and licenses when using Unitlab Annotate with the Free Plan.
NOTE! When you use Unitlab Annotate with the Free Plan, every dataset you release will be made public on Dataset page of Unitlab Annotate. Before releasing, you must agree to assign one of the following licenses to your dataset.
MIT
CC BY 4.0
Public Domain
BY-NC-SA 4.0
Discover Unitlab's Plans and Pricing to keep your datasets private by upgrading your plan.
In Unitlab Annotate, we support the following common export formats. Other formats are coming in the next version of Unitlab Annotate in the near future.
COCO
YOLOv5
YOLOv8
In our example, let's select the COCO format and click on the "Release" button. This creates a dataset based on the latest annotated data of your project.
As your annotations are corrected or updated over time, you can create a new version of the dataset based on the most recent annotations. You never lose any data; you can restore or download older versions of your datasets.
Each new release creates a new version of the dataset. You can create an unlimited number of versions of your datasets from your active projects, keeping everything organized and lossless.
In our demo, we currently have only one version of the dataset. Let's proceed to update the annotations, label all previously unannotated data, and create a new release with each update. We can add additional data and release other formats of our dataset as well:
In this demo, we created 4 versions of the project's dataset. To view all releases of your project, you can navigate to either this page or the Dataset page of Unitlab Annotate.
Let's find the dataset we have just released on the Datasets page of Unitlab Annotate, as shown below in the screenshots.
On the Dataset page of Unitlab Annotate, you can control the versions of the datasets that are released from projects.
