A bounding box is a rectangle positioned on the object.
To annotate using a bounding box, select an object class and then do one of the following:
- Click and drag the cursor on the asset and then release the cursor.
- Select annotation vertices with two mouse clicks. The first click marks the upper-left vertex of the rectangle and the second click marks the lower-right vertex.
Bounding boxes can be:
- Used in conjunction with tasks like Object/entity detection, Transcription and Classification.
- Nested with Classification and Transcription tasks.
In Image projects, when you copy&paste an object (for example a bounding box), the contents of its sub-jobs, if any, are also copied. This can help speed up labeling similar complex objects.
For PDF-related jobs:
It is not possible to make named entity recognition (NER) annotations where a bounding box has already been made. In this situation you can:
- First make NER annotations, and then add the bounding box annotations above the exiting NER annotations.
- Using the job viewer, hide specific bounding boxes and then annotate named entities.
For information on how to use the bounding box tool with video projects, refer to Labeling video assets.
Updated about 23 hours ago