After a labeling job is created, job packages move through labeling, review, and acceptance before the labeled data can be used for model training. This topic walks through the full workflow using image labeling as an example. The steps for other labeling types are similar.
Prerequisites
Before you begin, make sure that:
Job packages have been distributed by an administrator. For details, see Create a labeling job.
Your account has the workspace administrator or labeling administrator role. If not, contact a workspace administrator to assign the labeling administrator role to your account. For details, see Manage members of a workspace.
How it works
Each job package moves through three stages before the labeled data is ready for model training:
| Stage | Who acts | What happens |
|---|---|---|
| Label task | Labeler | Claims the job package, labels all images, and submits. |
| Quality inspection task | Reviewer | Reviews submitted labels and approves, modifies, or rejects them following on-screen instructions. |
| Acceptance task | Acceptance admin | Makes the final acceptance decision by accepting, modifying, or rejecting the labels following on-screen instructions. |
Open iTAG
Log on to the .
In the left-side navigation pane, click Workspaces. On the Workspaces page, click the name of your workspace.
In the left-side navigation pane, choose Data Preparation > iTAG.
On the iTAG page, click Go to the iTAG Page in the upper-right corner.
Claim and label a job package
On the iTAG platform tasks page, select your workspace in the upper-right corner.
In the left-side navigation pane, click Label Task, Quality Inspection Task, or Acceptance Task to view job packages of that type.
In the Assignment Available section, find the job package to claim and click Grab the order immediately. After you successfully claim the package, the labeling page opens.
On the labeling page, use the labeling tools to label each image. All images in the package appear as thumbnails in the left-side pane — click a thumbnail to switch to that image.
After labeling all images, click Submit in the upper-right corner.
Review labeling results
In the left-side navigation pane of the tasks page, click Review Jobs.
Review, modify, or reject the labeling results in the job packages that you have claimed by following the on-screen instructions.
Accept labeling results
In the left-side navigation pane of the tasks page, click Job Acceptance.
Accept, modify, or reject the labeling results in the job packages that you have claimed by following the on-screen instructions.
What's next
Export the labeled data from completed job packages and use it to train models. For details, see Export labeling results.