Previously, when creating a Job, it was straightforward to do: we take the selected frames, the name and the assignee, create the job and notify the assigned party about the task. However, with the introduction of the Autolabelling, a Job can also become an AI model enabled Job, resulting in a few changes to the flow: [Job Creation Image Needed Here]

When the Auto Label button is selected, a playground popup appears, where the user can test and adjust label configurations on a subset of frames. The user can select suitable labels for the frames of interest, add descriptions and a confidence score for each label, and then test these configurations on multiple images in order to decide which confidence levels and descriptions give the best result. [Job Creation Auto Label Playground Image Needed Here]

Once the user decides on the optimal configuration, the job is created and the auto-labelling process starts in the background. While you can navigate away to complete other tasks, the selected model generates the appropriate annotations in accordance with the confidence score provided per label. [ Auto Label Job with Labelling and completed status Image Needed Here]

While we are confident on the performance of our flows and incorporated models, these annotations are best subject to review, for predictions are only as good as the confidence score and descriptions that are provided to labels of interest. It is always best to have a human review and adjust the automatically generated annotations in order to ensure quality.