- Create labels "good" and "bad".
- Create a task with those labels and the raw dataset.
Optionally, choose tag type instance
Optionally, have it copy to a a new dataset called "filtered"
- Create an Action to move only data with the "good" tag into a new dataset "good_only".
Or, wait until the job completes, and then query the data once and move it into a new set.
Now you will have all the data in a "good_only" dataset.
- Create a second task that watches the "good_only" set for the main labeling work.
Updated about 2 years ago