Auto-Label
Auto-Label is one of the most powerful features of the Suite. It provides an AI-based automated labeling function that automatically detects and labels objects, as well as helps automate the quality assurance process in data labeling.
Why should I use Auto-Label?
In addition to simply automatically labeling data, the Auto-Label helps accelerate data labeling and quality assurance in two additional ways.
Measuring and classifying labeling task difficulty
When Auto-Label is applied, it not only automatically labels data but also measures the difficulty of the labeling task. The estimated difficulty is displayed in the Difficulty column of the Label table. (*red=hard, yellow=moderate, green=easy).
Users can easily sort and filter labels by difficulty.
Highlighting annotations with high uncertainty
Among the annotations created using Auto-Label, high uncertainty annotations are highlighted in yellow in Review Mode. The high uncertainty could be due to anything such as occlusions, small object size, bad lighting conditions, complex scenes, and so on. Users can focus on reviewing and editing these annotations highlighted by AI.
Configure up Auto-Label
To execute Auto-Label, you first need to link each object and the Auto-Label AI.
- Go to the “Auto-Label Settings” tab in the sidebar of the project page.
- 1.Auto-Label Settings
- Map each object class and Custom Auto-Label AI on the Auto-Label Settings tab, and complete the Project Configuration (Configure Auto-Label).
2. On the “Configure Auto-Labeling” window, please follow Step 1~3 in order.
- Step 1. Select object classes or categories that you want to set Auto-Label.
- Step 2. Select an Auto-Label model to link to the selected object class.
- Step 3. Select detailed object classes (multiple selections are allowed).
3. Click the Save button on the bottom right corner to finish configuring the Auto-Label.
Run Auto-Label
1. Go to the Labels tab in the sidebar of the project page.
2. In the Label table, select the label(s) you want to apply Auto-Label to.
3. Click the
Auto-Label button and the Auto-Label AI will start labeling.4. Refresh the page from time to time to check whether Auto-Label has been completed.
- When Auto-Label is in progress, a blue icon is displayed in the Pre-Label column of the Label table.
- When Auto-Label is completed, the estimated difficulty level is displayed along with a blue icon in the Pre-Label column.
Additional Notes
Supported Annotation Types
Bounding Box
Common Objects
Polygon Segmentation
Common Objects
Keypoint
'Human' keypoint template
* Currently, only the "person" class is supported and you must set the annotation type as "Keypoint - Human template".
Polyline
Not Supported
❗ Please refer to the following manual for support information for each annotation type of 'Custom Auto-Labeling'.
Supported Usage Plans
The Auto-Label feature is available for all plans, and the amount of Auto-Label credits available for each plan varies.
Auto-Label Credits
Auto-Label credits are deducted when you apply Auto-Label. Please refer to the relevant manual for details.
Video Auto-Label
- For Video projects, Auto-Label also includes tracking of objects throughout the video. Auto-label detects changes in objects in consecutive frames and automatically assigns matching object IDs.
- When you use Auto-Label in your video project, you will be charged based on the total number of frames, not the number of videos, to which Auto-Label is applied to.
FAQ
When I use all my Auto-Label credits, will Auto-Label no longer run?
This varies by your usage plan:
- In the Team and Enterprise plans, you can go over your monthly allocated Auto-Label credits, and you will be invoiced for the incurred overages.
- In the Free plan, Auto-Label will not be executed when all allocated monthly Auto-Label credits are used.
Last modified 1mo ago