Auto-Labeling
Auto-labeling is one of the most powerful features used in Suite. This feature allows you to label objects automatically so that you can effectively reduce the time for labeling and reviewing, which can ultimately lead you to get to production-ready training data faster. Hence, it is important to configure this feature when you create a new project. Please refer to the following to see how you can use the auto-label feature. (See: Configure Auto-Label)
You can use the auto-label feature by selecting the labels and clicking the “AUTO LABEL” button from the Label List page. When you click the “AUTO LABEL” button, the auto-label AI will automatically label objects for you.
The completed tasks are marked with blue icons in the “Pre-label” column.
In addition, the auto-label AI will process following tasks in order make your data labeling and reviewing process more efficient:
Measure and classify the difficulty level of each labeling task. Difficulty level is displayed on the “difficulty” column. (red → hard, yellow → medium, green → easy) You can sort and filter labels by its difficulty level.
Highlight annotations with high uncertainty values by marking them with a thick yellow bounding box so that a user can prioritize on reviewing these annotations.
The auto-label quota is set based on your plan, which you can check from the My Account Page(“Billing & Usage” tab).
For projects that contain Video data, our Auto-Label will also provide object tracking information. The Auto-Label AI will detect and track objects throughout the video and automatically provide object IDs.
Please note that when applying Auto-Label to videos, you'll be metered based on the number of frames Auto-Label is applied to.
