Log In

Object/entity detection jobs

To detect objects in an asset, use our object labeling tools:

Available labeling tools

  • Point
  • Line
  • Polygon
  • Bounding box
  • Semantic segmentation
  • Vector

Availability of object detection labeling tools depends on your project asset type. Refer to this table:

Labeling toolImageTextPDFVideoFrame
Bounding box
Semantic segmentation

For information on detecting objects in plain text, refer to Named entities recognition.

For additional information regarding some of the specific object detection tools, refer to:

Using labeling shortcuts speeds up your work. To access the full list of labeling shortcuts:

  • Press Alt/Option + S
  • Click on the shortcut icon.

Before you start labeling, make sure that you configured your interface correctly. Refer to Customizing project interface.

Named entities recognition (NER)

With Kili, you can do named entity recognition (NER) for text and PDF type assets.

You can select entities:

  • at character level
  • at token level by double clicking on the token (text only)
  • across phrases
  • across paragraphs
  • overlapping a previous selection


To simplify annotation, we process tokens to remove extra line returns/extra spaces at the end of the token. No need to post-process them.

Plain text annotation example:

If you want to rapidly select all occurrences of an entity in the text, right click on the entity and select "Propagate".


PDF annotation example:

You can also annotate plain text with structured style (rich text) with:

  • bold/italic/underline text
  • borders
  • paragraphs
  • different font sizes
  • margins
  • tables
  • everything else that plain HTML enables


Importing rich text is currently only supported through the Kili API. Refer to an example tutorial


When finished with Named Entity Recognition, you can perform another related task called Named Entities Relation. For details, refer to Object/entity relation jobs.

Did this page help you?