Processing a single data point can involve multiple tasks. For example, if you have a video bounding box project and submit one data point in the form of a 30-second video, each labeler might only add a single bounding box to 1 second of video, even though the output shows 100 bounding boxes across the 30-second clip. In this case, each bounding box counts as one task.
Viewing information at the task level tells you how many tasks there were and who processed them. You can also inspect task outputs for mistakes. Task information can help you to identify bottlenecks in your data pipeline.
- Head to your super.AI dashboard
- Open the relevant project to bring up its work queue
- Open a data point by clicking on its row in the work queue
- Click on Tasks at the bottom left of the data point details view to expand the task information
- Click on a task in the table to view the task. The task view shows the input and output for that task, as well as the instructions, the labeler's email (if you're using your own labelers), and the status of the task (e.g.,
Complete). The task view is view only; you cannot edit the output of a task. From this view, you can cycle through tasks using the arrows above the input and output boxes. Tasks are numbered in chronological order.
For privacy reasons, the identity of the labeler is only available if you use your own labelers
Updated over 1 year ago