Skip to main content
How to Export Annotation CSV File

How to export detailed insight about the annotation progress on your project

T
Written by Taylor Maggos
Updated over 3 months ago

Think of the annotation CSV file as an annotation feedback report, which is just what we named it!

Whether you are done with annotation or just midway through the project, you might want to see some more insight to how annotation is progressing. This is especially helpful if you have a data labeling team, and you want to monitor their work.

In another article we go more in-depth about workflows, tasks, and phases, BUT if you haven't read that here is the synopsis:

  • When you upload media (image or video) to a dataset, each piece of media gets a unique media ID, if you transform that media, new media IDs are created for each new piece of media, and they are tied to the parent media ID-- so we always know where everything came from.

  • When you attach that dataset to a project, each piece of media is transformed into a task. Each task has its own Task ID to help us track everything along the way.

  • Each task in the project is then broken down into 3 unique workflow IDs, because there are 3 phases of labeling/Quality assurance.

    • Therefore, the progression goes, 1 media ID --> 1 Task ID --> 3 workflow IDs

This reminder will be helpful as you look at your annotation feedback report CSV file, because these names --workflow ID, Task ID, Media ID -- come up!

How to download the CSV file:

  1. Navigate to your Project, and into the Annotation Tab.

  2. Scroll down to the Annotation workflows section and you should see a button which says β€œDownload CSV”

  3. Click Download CSV and the file will automatically download to your computer!

What you will find in the CSV File:

  • Workflow ID (the unique ID given to each task per phase, each piece of media will have a task ID and then 3 workflow IDs – 1 for each phase of annotation)

  • Media ID (each piece of media you upload will have a unique ID)

  • Dataset ID (the dataset ID the piece of media came from)

  • Task Status (is the task submitted, assigned to someone to label, rejected, skipped, etc.)

  • Task URL (the URL link to the specific phase of the task, ex. Phase 2 task link)

  • Phase (image_boxing = Phase 1, Image_boxing_review = Phase 2, Final_Review = Phase 3)

  • Annotator Name (who labeled the specific phase of each task)

  • Time to Complete (time, in minuets it took for the annotator to label that phase)

  • Project Name (This should be the same for the whole CSV file you download, the project name)

  • TimeSecs (the time in seconds it took the annotator to label that phase)

  • Completed At (what date and time the phase was labeled)

  • Frame Count (if video there will be the number of frames per task here, if images this column will be blank)

Did this answer your question?