🖌️Annotation
Annotate images for your computer vision projects
Last updated
Annotate images for your computer vision projects
Last updated
In this section, we will learn how to manage projects, annotate data, and export the annotated data in Unitlab Annotate. To start annotating, you are required to set up a project. If you have already created a project, you can begin annotating right now.
Learn how to create a project instantly on the Setup a Project page.
In the previous section, we created a Vegetable and Fruit Segmentation Project for demo purposes. Let's continue with this project in this section to learn how to annotate data.
To start data annotation, simply click the "Start Annotating" button in the right corner of Unitlab Annotate.
When you click, It leads you to the annotation tool of Unitlab Annotate. As we selected the Image segmentation annotation type for our demo project, Unitlab Annotate's UI automatically adapts to the segmentation tasks.
Unitlab Annotate is powered by several auto-annotation tools. Learn how to use them in the Auto-Labeling section.
The auto-annotating tools of Unitlab Annotate are amazing! They can speed up the manual annotation process by up to 15 times. All auto-annotation tools provided by Unitlab Annotate are free, with no subscription required. We encourage active usage of these tools and recommend refining their results through human review to obtain high-quality datasets.
Learn how to use the Auto-Annotation tools of Unitlab Annotate on the following pages.
The annotation tools in Unitlab Annotate are dynamic and vary based on the annotation type. For example, the tools used in Image Bounding Box annotation differ from those in Image Segmentation annotation.
In Unitlab Annotate, histories track every action for each annotation. This feature is crucial for monitoring and understanding progress in data annotation.
You can easily view each annotation history on the History page of Unitlab Annotate. The annotation histories can be categorized as follows:
Seen: Viewed by a Member
Labeled: Labeled by the Annotator Position
Reviewed: Reviewed without any change by the Reviewer Position
Fixed: Reviewed with a change by the Reviewer Position
Learn about Workspace Member roles, as well as Annotator and Reviewer positions, on the Role-based Access page.
Commenting on annotations is an important feature in Unitlab Annotate, as comments are referred to as 'Issues' and need to be resolved. It enables reviewers to create issues by commenting on the annotations.
As comments are referred to as 'issues', we can view the list of issues on the project page of Unitlab Annotate as shown below.
You receive a notification when your annotated data is reviewed or fixed, and members who reported issues will also be notified when they receive a reply. All actions related to issues are consistently communicated to the assigned members, ensuring effective monitoring and timely resolution of each issue.
Let's annotate data using the auto-labeling tools of Unitlab Annotate. There are several auto-annotating tools available in Unitlab Annotate. Let's use the Magic Touch , which is powered by the Segment Anything Model (SAM).
Check out the demo video below, which demonstrates how the Magic Touch efficiently and accurately annotates data in the demo project.
.svg?alt=media&token=f999c652-c1b6-4b51-8cc0-d2d899b6f634){kind=small}