Auto Annotation is the process of creating labels for a dataset using the reference model.
The training weights from any training session of an experiment can be used to auto annotate.
A dataset with number of images and an empty annotation set must exist before proceeding.
To auto annotate go to the datasets page and click on the "Auto Task" button described by the following icon.
- Click on "Add Task".
- Select "Auto Annotation".
- Add a Name to the auto annotation task.
- Model Location - Select From experiment (Custom weights file upload is not support at this time).
- Select the Experiment - A model training session must be created before auto annotation).
- Select the weights from a training session (best weights in this session will be used).
- Finally select the destination dataset and the target annotation set in this dataset.
- Click "create task".
- The task process will be shown in a card for this task.