Projects

Understand projects and how they connect to other parts of the Labelbox platform.

A project is a container that houses all of your labeling operations for a specific set of data rows.

Ideally, each project is highly customized to create a specifically labeled dataset. For example, you may create a separate project for each of the following situations:

  • You've identified an edge case that your model struggles to predict and you only need to label the data rows that contain that specific scenario.

  • You need a large amount of training data and it is too much work for one labeling team. You can create separate projects that reference the same ontology and distribute the labeling work across multiple teams.

  • You have highly specialized labeling teams that are trained to annotate specific data modalities (e.g., images and text). In this case, you can create separate projects and tailor the ontology to label those data types specifically.

The project is where the ontology, the data rows, and the editor interface connect. Projects are designed to make your labeling operations as efficient as possible.

Create a project

To create a project, please view the dedicated documentation page here.

Project overview tab

The Overview is the first tab that you land on when you click on any project. This tab is designed to provide a quick glance at the most important statistics of your project.

The overview tab is divided into three parts:

  • Labeling progress: This section provides a list of all statuses in a project and a count of data rows in each status. It also provides a count of the issues.
  • Labeling and workflow tasks: This section lists all the tasks in the project, including the Initial labeling task, the Rework (all rejected) task, and all review steps.
  • Analytics view for annotations: This section lists all the features in a project along with the count of annotations for each.

Labeling progress

This section provides a simplified set of counts based on the set of statuses in a project. These statuses are a summary view reflective of a project's workflow and provide general insight into the project's state. See the documentation on Workflows for more information.


Below is a description of each bucket displayed in the Labeling progress section. The counts will always be consistent between the overview tab and the left-side summary of the data rows tab.

StatusDescription
To labelData rows that don't have all of their required labels yet (this excludes skipped data rows).
In reviewData rows that have been labeled.
In reworkData rows that have been labeled or skipped and currently are in the rework task.
SkippedData rows that have been skipped.
DoneData rows that have passed through all review and rework stages in the workflow.
IssuesCount of total issues in the project.

Labeling and workflow tasks

This section provides a list of all tasks in the project and and associated count of data rows currently in the task.


Below is a description of the three tasks that exist out of the box in each newly-created project. Reviewers can enter the rework task or a specific review task directly from the overview tab using the buttons displayed above.

Task nameDescription
Initial labeling taskTask reserved for all data rows that are in the process of being labeled.
ReworkTask reserved for all data rows that are in the process of being reworked, either after being rejected or explicitly moved to rework from another task.
Initial review taskTask reserved for data rows that have been submitted and need review.

Analytics view for annotations

This view provides the prevalence of different features within the project across data rows.

For each feature, you can see the following:

  • Name of the feature.
  • Count of data rows that contain the feature.
  • % Share indicating the percentage of data rows that contain the feature.
  • A horizontal bar chart that depicts the percent share. You can also click the bar to navigate to a filtered view of the data rows tab including only the data rows that contain the relevant feature.