AnnoStation
Sign UpAboutTermsPrivacy
English
English
  • AnnoStation Knowledge Base
  • Getting Started
    • Introduction to AnnoStation
    • Quick start
  • Mastering AnnoStation
    • Organization
      • Member types
    • Setting up a Project
      • Adding datasets
        • Importing images
        • Connecting to external data
      • Organizing datasets
      • Defining labels
      • Adding Project members
        • Project member roles
      • Assigning annotators
    • Working with a Project
      • Selecting target files to perform actions
      • Searching for images
      • Sorting and Filtering
      • Report panel
    • Workflow and status management
    • Annotating an image
      • Annotation tools
    • Reviewing annotation
    • Collaboration
      • Comments
      • Sharing Projects with external Organization
        • Project member roles in shared Projects
        • Workflow in shared Projects
    • Exporting datasets
  • Resources
    • JSON format
      • Annotation
      • Examples
    • Keyboard shortcuts
    • FAQ
Powered by GitBook
On this page
  • Overview
  • Create a new label
  • Custom attributes
  • Modify Labels
  • Delete Labels

Was this helpful?

  1. Mastering AnnoStation
  2. Setting up a Project

Defining labels

PreviousOrganizing datasetsNextAdding Project members

Last updated 4 years ago

Was this helpful?

Overview

Creating high-quality training data requires clean and well-thought labels.

AnnoStation supports the following two types of labels:

Type

Description

Segmentation

Segmentation is typically used to train the model to recognize the precise shape of an object. It is used to label each pixel of an image with a corresponding class of what is being represented.

Vector

Vector is often used to detect a certain object within an image. It is used to label the outline of shapes of an object with a corresponding class of what is being represented.

AnnoStation supports the following types of vector objects:

  • Point

  • Bounding Box

  • Polygon

  • Path

For vector labels, you can also define for additional classification.

Create a new label

  1. Go to the Annotation Labels tab.

  2. Click on Add label.

  3. Provide a name and choose a color.

  4. Select if the label is Segmentation or Vector.

    • For vector labels, you can define custom attributes.

  5. Select Create Label.

Custom attributes

When you want to classify an annotated object, using custom attributes is a powerful way.

AnnoStation supports the following types of custom attributes:

Type

Description

Single Text

  • Annotators classify an object using free texts.

Radio button

  • Annotators choose between a set of options.

  • Annotators can choose one option.

  • You can include "Other" as an option, and Annotators can type a short answer.

Checkbox

  • Annotators choose between a set of options.

  • Annotators can choose more than one option.

  • You can include "Other" as an option, and Annotators can type a short answer.

Dropdown

  • Annotators choose between a set of options.

  • Annotators can choose one option.

  • You can include "Other" as an option, and Annotators can type a short answer.

To add custom attributes:

  1. Against vector labels, select Add attribute.

  2. Provide a name of the attribute, and select the type.

  3. Add response options (if applicable for your attribute type).

  4. Include "Other" as an option if desired by selecting Write-in Choice.

  5. To add more attribute, click Add attribute.

Modify Labels

You can modify labels, however make sure you are certain about the action as it will impact all existing annotation associated with the label.

To modify a label, go to the label you want to modify, and select the pencil icon on the left.

  • Changing the name or color will update the existing annotations.

  • Deleting attributes or options will remove existing attributes or values.

  • Changing the type of an attribute may result in removing existing values if it is not possible to preserve.

Delete Labels

You can delete labels, however make sure you are certain about the action as it will delete all existing annotations associated with the label and it cannot be reversed.

To delete a label, go to the label you want to delete, and select the trash bin icon on the left.

custom attributes