Setup a Project
Get started with data annotation instantly
Last updated
Was this helpful?
Get started with data annotation instantly
Last updated
Was this helpful?
This step-by-step guide will show you how to build a dataset from scratch using Unitlab Annotate. You will learn how to create a project. Our instructions are easy to follow, even if you are new to data annotation.
To create a project in Unitlab Annotate, make sure you have created an account and are logged in. If you don't have an account yet, start by registering . After signing in, you are required to create a workspace. Learn how to create and manage workspaces on the page.
On the page, you'll find detailed instructions and helpful tips on how to set up your workspace effectively. Additionally, the page offers guidance on managing multiple workspaces, allowing you to seamlessly switch between different projects and tasks.
In the guide, we will learn how to set up projects and start annotating your data.
Let's get started!
When you visit Unitlab Annotate, there are three options available for creating projects:
Clone a quick demo
Create a project from scratch.
Clone a public dataset from the Dataset page.
In Unitlab Annotate, the list of demo projects are prepared for you to get started instantly. After clicking the "Clone Demo" button, a project will be created for you. You can then manage, edit, and test data annotation within it.
To create a new project from scratch in Unitlab Annotate, click the 'Add a Project' box on the project page. In this guide, we will be creating a Vegetable and Fruit Segmentation Project.
To successfully create a project in Unitlab Annotate, several steps need to be followed:
Project details
Upload data
Add classes
Add members
Confirm project
In the project details, the Project Name and Annotation Type are required, whereas choosing Auto-Labeling, writing Project Description, and uploading Project Description files are optional.
In our example, we need to give a name for our demo project and choose the Image Segmentation annotation type. Let's consider the other options as optional.
The next step is to upload data to the project you are creating. There are three methods available for data upload:
Upload data in the Unitlab Annotate Dashboard
Upload data through the CLI.
Upload data using our Python SDK.
Let's upload data directly from the local storage within the Unitlab Annotate Dashboard. We have prepared 5 images for this demo.
After uploading the data, simply click the "Next" button to proceed to the next steps.
According to your project's requirements, you need to add a class or a group of classes for annotating your data.
In our demo, we have selected common vegetable and fruit names as classes, as shown in the screenshot. You are free to choose class colors as per your preference.
Two options are available for adding members to Unitlab Annotate:
Hire Annotator - ⚡ Labeling Service
Invite Annotator
Unitlab offers an Integrated Labeling Service, where our team of expert annotators annotate your data for you. Please note, this option is available only with our paid subscriptions.
In the second option, you can either invite an annotator or assign yourself as an annotator in a project. In the demo, let's assign ourselves as the annotator.
This is the final step, where you should review and confirm all project details before proceeding. Once the project is created, you will receive an email regarding its creation, after which you can begin annotating your data.
After confirmation, you can find the project in the Unitlab Annotate project list as shown in the screenshot below.
The project has been successfully created; now, we can start annotating data. Let's learn how to annotate data in the next section.
If you have a large volume of data, we recommend uploading it using our CLI or Python SDK. You can find instructions on how to use them on the and pages.
The images can be in any format, such as PNG or JPG. When uploading from local to the web, you can upload up to 100 images per attempt. There is no rate limit if you use the or .
