📋 Kanban Board Overview

The Kanban board provides an intuitive, visual way to manage annotation progress across your entire team. It organizes your work into columns that represent different stages of the annotation process, with cards to show groups of frames moving through these stages.

Board Structure

The Kanban board consists of five main columns, each representing a distinct phase in the annotation pipeline:

ColumnPurpose
BatchesAvailable image batches ready to be split up for annotation
AnnotatingActive annotation jobs (manual and automated)
ReviewCompleted annotation jobs awaiting review
DatasetWhere approved frames are centralized and organized into train, valid, and test sets
RejectedAnnotations that have been rejected from the dataset

Card Workflow

1. Batch Cards

Located in the first column, batch cards represent collections of images ready for annotation. Clicking a batch card opens a preview page where you can:

  • View individual frames in the batch
  • Access the job creation sidebar, where you can choose between creating an automated or manual job and assign it to a teammate.

2. Job Cards

When you create a job from a batch, a new card appears in the “Annotating” column. Job cards come in two varieties:

  • ⏳ Autolabel Jobs: Display a spinner animation and “autolabelling” status

  • ✏️ Manual Jobs: Show assignee information and are clickeable, bringing up a details view and access to the Canvas

3. Review Cards

  • Autolabelling jobs will automatically move to the review column once complete
  • For manual jobs, the annotator submits the job for the review and it moves to the “Review” column.

In the review state, a different person can review the annotations for quality before adding them to the dataset

4. Dataset Cards

After successful review and dataset integration, cards move to the final “Dataset” column (or the “Rejected” column if they are rejected).

The dataset is available to view in the “Dataset” tab, but the cards in the column on the Kanban board show which jobs comprise the current dataset.

The Kanban board’s visual workflow makes it easy to track progress and identify bottlenecks in your annotation pipeline.