Label Editor - Annotation

This documentation provides a comprehensive guide to using the Label Editor for annotation, covering its components, features, and the display of various elements.

For information on how to label, see our documentation here.

​

Annotation Mode

​

Introductory Video Tutorials

The following video goes through the basics of annotating images in the Label Editor.

The following video goes through the basics of annotating videos in the Label Editor.

Annotation mode components:

• A. Editor menu
• B. Quick toolbar
• C. Classes
• D. Instances and Frame labels
• E. Automated labeling
• F. Editor timeline (for videos and image sequences only)
• G. Annotation canvas: displays the media asset and provides interface for drawing and editing labels.
• H. Editor header
• I. Skip or Submit

​

Copy / Paste Labels & Instances

Using the copy and paste features in the Label Editor can significantly save time and effort. Here’s how you can efficiently use these features:

• Copy: To copy both the label and the instance, use the keyboard shortcut CMD + C on Mac or Ctrl + C on Windows.

• Paste instance: To paste the copied instance, use CMD + V on Mac or Ctrl + V on Windows. Instances can only be pasted onto a different frame from where they were copied.

• Paste Label: To paste just the label / shape, use CMD + Shift + V on Mac or Ctrl + Shift + V on Windows. Labels can be pasted onto any frame.

​

Bulk Label Operations

Fixing incorrect labels is an important step in producing high-quality datasets. With bulk label operations, annotators can merge labels which should really be one instance, or efficiently delete multiple labels in bulk.

Bulk label operations are accessed in the editor menu.

Merge operations

Merge one or more label objects into a single object.

1. Select all objects to be merged in the Select objects to be merged drop-down menu.
2. Select the target object in the Select objects to be merged into field.
3. Click Merge.

Delete operations

Encord lets you perform bulk delete operations as follows:

Remove labels of a specific class:

1. Select a class to delete from the Select class menu.
2. Specify the frame range.
3. Click the Delete button.

Delete all labels within a frame range:

1. Navigate to the Delete all labels from range section.
2. Set a range to delete labels from.
3. Click Delete.

Delete labels below a specified confidence level:

1. Navigate to the Delete labels below confidence section.
2. Set the minimum confidence threshold. All labels with a confidence value below this threshold are deleted.
3. Specify a range of frames to delete the specified labels from.
4. Click Delete.

​

Classes

In annotation mode, the Classes section of the Label Editor shows you the available Ontology classes, both objects and classifications. You can create a new instance by either clicking the desired class, or by using the indicated instantiation hotkeys. Learn more about Ontologies in general here or learn more about the details of Ontology structure on Encord.

​

Instances and Frame labels

The canvas is your labeling area to annotate the instances of the Ontology classes.

• An instance is a specific occurrence of a class. For example, if you have a class called “Car”, an instance of this class could be “Car(0)”, which might represent a specific black sedan. This single instance can appear in a single frame or a range of frames, and therefore, instances may contain multiple labels across frames. You can think of an instance as a unique object that belongs to a certain class.

• A frame label is a frame-specific annotation of an instance. For example, the annotation of “Car(0)” on frame 201 is a label. Labels are used to annotate instances in specific frames, providing more detailed information about the instance in the context of that frame.

​

Skip or Submit labels

Submit tasks

Labels are applied in the Label Editor. Click the Submit button to submit your labels for review.

Skip tasks

Users have the option to skip tasks in cases where, for example, the file is blurred, or if they suspect the task was erroneously added.

When a user skips a task, it is automatically unassigned from them, and the task’s status is updated to Skipped. If another task is available, it automatically opens in the Label Editor and is assigned to the user.

1. Click the Skip button to skip an annotation task.

2. A dialog appears. Enter a reason for skipping the task - the reason is added as an issue to the task that annotators and reviewers can see. If no reason is entered “This task was skipped” is added by default.

​

Turbo mode

Turbo mode simplifies the standard annotation workflow, enabling annotators to rapidly assign or confirm attributes. Activate turbo mode and choose an instance from the top-left dropdown menu to set the instance and its relevant attributes.

Enable turbo mode:

1. Open the editor settings.
2. Use the search bar to search for Turbo mode.
3. Enable the Turbo mode toggle.

​

View Annotator Instructions

The Annotator instructions button is available in the top-left corner of the Label Editor if annotation instructions are provided by the Admin or Team Manager.

​

Confidence Score

Each instance or label is assigned a confidence score (denoted with the symbol α) that appears next to the label name.

The confidence score is a measure of a machine learning model’s certainty that a given prediction is accurate. The higher the confidence score, the more certain a model is about its prediction.

Manual labels are always assigned α = 100%, while label predictions created via models and automated methods such as interpolation will have a confidence score below 100% (α < 100%).

Models allow you to select a Minimum confidence threshold that determines the lowest confidence that label predictions can have to appear as annotations. It ranges from 0 to 1.

A Minimum confidence threshold of 0.6, as seen in the image above, states that only labels above a confidence score α = 60% will be accepted as labels.

​

Labels pane

The Labels pane shows all the instantiated objects and classifications for a data unit. Objects are grouped first by class, and sorted by creation time within each class, with earlier objects on top. They are named in an “ABC (0)” format, where ABC is the Ontology class and 0 is the first instance annotation.

• A: Toggle between showing displaying all labels, or only labels in the current frame.
• B: Show / Hide details for this instance.
• C: The range of frames the instances appears on.
• D: Attributes the object has and the frames they appear on.
• E: Filter by frame range.
• F: Sort label instances.
• G: Search for an instance.
• H: Show / Hide labels.
• I: Show / Hide labels for this instance.
• J: Additional actions. Described below.
• K: Keyboard shortcut to create a label for this instance.
• L:Edit attributes for this instance.
• M: Add a new label for this instance.
• N: Toggle auto-annotation for SAM.

Click the three dots icon to access additional actions:

Copy URL: Copies the URL for the particular object.

Copy identifier: Each object instance is automatically assigned a unique alphanumeric ID, which can be used to identify it within the project. This ID is sometimes known alternately as a feature hash, instance hash, or instance identifier. For objects and classifications, it may also be known as the object hash or classification hash. Use this button to copy the instance identifier to your clipboard — which may help in workflows using specific instances in source code or for other purposes.

Track from this frame: Initiates single object tracking using SAM 2. This type of automated labeling creates labels for the specified object from the current frame, according to the Single object tracking settings section in the editor settings.

Interpolate: Interpolates this object’s location for a range between specified frames. See our documentation on interpolation for more details.

Go to object: Navigate to the specified object in the data asset. If it is present in more than one frame, the first frame the object appears is shown.

Split object track: In some cases, particularly when many objects overlap, labels may be associated with the correct instance in several frames, but later become associated with a different instance. In those cases, use the Split object track feature, which will split the labels from one instance into two at the frame specified.

Change class: Use the drop-down in the modal to change the current class of an instance. You can only switch within classes of the same annotation type. For example, you cannot switch from a bounding box to a polygon.

Delete object: Delete an object from a single frame or a range of frames. To delete labels between frame ranges, move the frame slider to the desired frames, or input the From and To (exclusive) values in the relevant form.

​

Attributes

The details pane allows you to modify attributes efficiently. For dynamic attributes, use the ‘Preserve chosen state’ feature. When this feature is toggled ‘ON’, any subsequent labels for that instance automatically inherit the same dynamic attribute value as the frame where ‘Preserve chosen state’ was enabled.

If you have numerous attributes, use the search bar to quickly find specific attributes or attribute values.

​

Automated Labeling

Automated labeling includes several techniques to quickly create labels automatically.

• Interpolation do not require set up or training, and are always available.
• SAM 2 Segmentation creates labels by clicking parts of your frame or image.
• SAM 2 tracking tracks objects across a range of frames using the [SAM 2] model.

1. Click the Automated labeling button to open the Automated labeling panel.
2. Select the type of automated labeling you want to use.
3. Run automated labeling for the selected objects.

Alternatively, run object tracking or interpolation on videos, image groups, and image sequences by:

1. Right-clicking a label.
2. Selecting Track from this frame to initiate object tracking, or select Interpolate to interpolate the selected object instance.

​

Edit Polygons and Polylines

Encord provides multiple tools to edit polygon and polyline labels after they have been created. See our tutorial video for an quick introduction on working with polygons in Encord.

​

Editor Canvas Interactions

Shape-specific interactions:

Right-click interactions:

​

Controls Common To Both Modes

​

Editor Layout

You can customize the default Label Editor layout by uploading a JSON file that defines your preferred arrangement. These JSON files can be added in the Project settings and are applied to all tasks in a Project.

Custom editor layouts depend on a file’s client_metadata or their DICOM tags. This means that custom editor layouts only work for files that contain client_metadata or DICOM tags. Watch the video tutorials below to learn how it fits together.

To upload an editor layout:

1. Navigate to the Project settings.
2. Click on Editor Layouts.
3. Click Upload JSON.
4. Select the JSON file containing the layout you want the Label Editor to have.

JSON file requirements:

The JSON file must follow the JSON schema defined here. The following fields are required:

Complete example

1

Add client metadata to files

Editor layouts are based on DICOM tags for DICOM files or on client_metadata for all other use cases. We provide templates for common mammography layouts based on DICOM tags here.

This example uses client_metadata. The following SDK script can be used to add client metadata to specific data units in a Dataset.

Copy

from encord import EncordUserClient
from encord.client import DatasetAccessSettings

# Instantiate Encord client by substituting the path to your private key
user_client = EncordUserClient.create_with_ssh_private_key(
                ssh_private_key_path="<private_key_path>",
            )


# Specify a Dataset to read or write metadata to
dataset = user_client.get_dataset("<dataset_hash>")

# Fetch the Dataset's metadata
dataset.set_access_settings(DatasetAccessSettings(fetch_client_metadata=True))

# Add your file names and client metadata here
layouts = {
  'Image1.jpg': {"encord-EditorGridPosition": "A", 'encord-LayoutGroup': '1'},
  'Image2.jpg': {"encord-EditorGridPosition": "B", 'encord-LayoutGroup': '1'},
  'Image3.jpg': {"encord-EditorGridPosition": "A", 'encord-LayoutGroup': "2"},
  'Image4.jpg': {"encord-EditorGridPosition": "B", 'encord-LayoutGroup': '2'},
  'Image5.jpg': {"encord-EditorGridPosition": "A", 'encord-LayoutGroup': '1'},
  'Image6.jpg': {"encord-EditorGridPosition": "B", 'encord-LayoutGroup': '1'},
  'Image7.jpg': {"encord-EditorGridPosition": "A", 'encord-LayoutGroup': "2"},
  'Image8.jpg': {"encord-EditorGridPosition": "B", 'encord-LayoutGroup': '2'},
  'Image9.jpg': {"encord-EditorGridPosition": "A", 'encord-LayoutGroup': "1"},
  'Image10.jpg': {"encord-EditorGridPosition": "B", 'encord-LayoutGroup': '1'},
  'Image11.jpg': {"encord-EditorGridPosition": "A", 'encord-LayoutGroup': "2"},
  'Image12.jpg': {"encord-EditorGridPosition": "B", 'encord-LayoutGroup': '2'},
}

for dr in dataset.data_rows:
  dr.client_metadata = layouts[dr['data_title']]
  print (dr.title)
  print (dr.client_metadata)
  dr.save()

2

Create a JSON file

Create a JSON for specifying the editor layout that suits your needs.

In the JSON file below:

• The grid arrangement is configured to display two files side by side in the Label Editor.

• The gridContent section specifies that tasks in either position (0 or 1) can have client metadata values for encord-EditorGridPosition set to either A or B.

• The topLevelGridFilter is defined as encord-LayoutGroup, meaning tasks with matching encord-LayoutGroup metadata values are displayed together in the Label Editor, ensuring they appear side by side when they share the same encord-LayoutGroup values.

  Copy

  {
    "grid": [
      [0],
      [1]
    ],
    "gridContent": {
      "0": {
        "client_metadata_query": {
          "or_conditions": [
            {
              "key": "encord-EditorGridPosition",
              "operator": "=",
              "value": "A"
            },
            {
              "key": "encord-EditorGridPosition",
              "operator": "=",
              "value": "B"
            }
          ]
        }
      },
      "1": {
        "client_metadata_query": {
          "or_conditions": [
            {
              "key": "encord-EditorGridPosition",
              "operator": "=",
              "value": "A"
            },
            {
              "key": "encord-EditorGridPosition",
              "operator": "=",
              "value": "B"
            }
          ]
        }
      }
    },
    "topLevelGridFilter": {
      "client_metadata": {
        "key": "encord-LayoutGroup",
        "operator": "="
      }
    },
    "$schema": "https://raw.githubusercontent.com/encord-team/Annotate/main/editor-layouts/editor_layout_schema.json"
  }
  

3

Create a JSON file

Upload the JSON file to your Project in Encord.

1. Navigate to the Project settings.
2. Click on Editor Layouts.
3. Click Upload JSON.
4. Select the JSON file containing the layout you want the Label Editor to have.

4

Create a JSON file

Open any task in the task queue of your Project.

The JSON file in this example results in the following Label Editor layout.

​

Editor Menu

Open the editor menu by clicking the menu icon in the top left corner of the Label Editor.

The editor menu contains the following controls:

• Return to Project - Returns the user to the Project navigation.
• Save changes - Saves the current state of all labels.
• View - Allows users to change what information is displayed next to all labels.
• Bulk label operations (annotation mode only) - Allows the user to perform actions on many labels at once.
• Automated labeling (annotation mode only) - Lets the user choose between several automated labeling techniques.
• Dark mode - Toggle dark mode on and off.
• Keyboard shortcuts - Displays a full list of keyboard shortcuts.
• Editor settings - Opens the Editor settings.

​

Editor Header

The Editor header at the top-right of the navigation bar contains buttons and menu items to perform labeling activities as well as manage general app settings and notifications.

Icon	Name	Description
	Help	Links to the: quick start guide, documentation, learning hub, and Encord support team.
	Notifications	Shows the current status and history of actions taken in the application. This includes uploading datasets, training models, and exporting labels.
	Share	Copies the URL of the asset currently being labeled or reviewed.
	Save labels	Manually saves labels on the Encord server.
	Label Editor settings	Settings controlling your labeling and reviewing experience.
	Issues	Shows any issues and comments for the current task, such as any objects or classifications marked as Required, or comments added by other users.

​

Quick Toolbar

A toolbar with handy tools to:

• Rotate the label editor.
• Adjust filters; contrast, brightness, gamma, pixelated scaling, and grayscaling.
• Remove polygon vertices.
• Add comments or issues to a frame or image.
• Quickly center an image.
• Zoom in and out.

​

Rotate the Label Editor

Click the Rotate toolbar icon on the quick toolbar to bring up a pop-up that enables you to rotate the label editor by using the slider, as shown below.

​

Pixelated Scaling

The Pixelated scaling feature enhances the visibility of individual pixels in an image or frame by disabling the browser’s automatic edge smoothing. This increased pixel clarity aids annotators in creating precise, pixel-perfect annotations, thereby improving their annotation accuracy, especially when zooming into an image.

Pixelated Scaling turned on	Pixelated Scaling turned off

Enable Pixelated scaling by selecting the Adjust filters icon ()in the Quick toolbar and enabling the Pixelated scaling checkbox, as shown below.

Pixelated scaling is a browser-specific feature. The table below shows which types of files Pixelated scaling works for Firefox and Chrome.

​

Grayscale Filter

The Grayscale filter feature enhances the visibility across an image or frame by applying an RGB grayscale to the image/frame. The filter applies across the pixels of an image/frame using the value for Red, Green, or Blue as a base then converting to grayscale.

Grayscale Image/Frame

Grayscale DICOM/NIfTi

​

Media Navigation Controls

The video frame slider lets you quickly scroll through frames in a video asset. Below the frame slider are a range of controls to let you play, skip and navigate through media. Notice that the navigation controls are not displayed in single images files, as well as any media with only 1 frame.

Icon descriptions:

The Time and Frame rate information for each frame is displayed in the far lower right corner.

​

Issues and Comments

Issues can be added to annotation and review tasks to communicate with other users. All issues contain comment threads relating to the entire task or specific sections of the task. A small notification badge on the Issues icon displays the number of unresolved issues in a task.

Types of Issues:

• Label rejection: Created when a reviewer rejects labels. The reviewer is prompted to provide a reason for the rejection in the form of a comment. If all instances in a frame or image are rejected, a separate rejection issue is created for each label. Labels with rejection issues display a red Issues icon next to the instance name, which turns green once the issue is resolved.

• File: Apply to the entire file and can be created at any stage within the Label Editor.

• Pinned: Apply to a specific part of the file or frame and can be created at any stage within the Label Editor. Red Pin icons indicate unresolved issues and remain red until they are marked as resolved.

• Frame: Apply to a specific frame and can be created at any stage within the Label Editor. Frame issues can only be created when multiple frames are present; such as in a video, image sequence, or DICOM volume.

• Frame range: Apply to a range of frames and can be created at any stage within the Label Editor. Frame range issues can only be created when multiple frames are present; such as in a video, image sequence, or DICOM volume. Frame range issues can only be created in Encord Active.

• Task requirement: Are created for each object or classification marked Required in the Ontology. The issue is resolved when the required object or classification is applied. Task requirement issues are the only issues visible in a Consensus annotate stage.

To Create a new Issue:

1. Open the Issues pane by clicking the Issues icon.
2. Click the New issue icon (or keyboard shortcut Ctrl+N).
3. Click the Apply to dropdown to select the type of issue you want to create.
4. Select any Issue Tags you want to apply to the issue.
5. Add a comment to the issue.
6. If you selected ‘Pin on canvass’ in Step 3, click the desired location on the canvas.
7. Click Create issue.

To Resolve an Issue:

1. Open the Issues pane by clicking the Issues icon.
2. Click the tick to mark a thread as resolved.

Resolved issues remain visible but can be filtered out. All users can resolve issues.

To Delete an Issue Thread:

1. Open the Issues pane by clicking the Issues icon.
2. Click the Ellipsis icon to show more options.
3. Click Delete thread.

To Reopen an Issue Thread:

1. Open the Issues pane by clicking the Issues icon.
2. Click Reopen on an issue that was previously resolved.

Use Cases:

• Ask questions: Annotators and reviewers can ask questions about the task. Admins can use this feedback to improve guidance.
• Explain task rejection: Reviewers can provide reasons for rejection, helping annotators avoid repeating mistakes.
• Track unresolved issues: Keep a record of open discussions that need follow-up before finalizing annotations.

​

Editor Lock

The ‘Editor lock’ is a safety feature we implemented to prevent several people making changes to the same file - annotators labeling, or reviewers reviewing the same data unit.

A warning is triggered in two cases:

1. When you have the same task open in two different tabs or browser windows:

2. When a different user is editing the same task:

​

WACOM Tablet Support

Our Label Editor fully supports Wacom tablets and pens. You get the best experience if you enable freehand drawing mode.

​

Label Editor V1 (Legacy)

The following actions vary significantly between the legacy Label Editor V1 and the new Label Editor. To see how to perform these in the new Label Editor, see our documentation here.

Adding a polygon vertex - Label Editor V1

To add more vertices to a polygon after it has been created:

1. Click the polygon you want to add a vertex to
2. Click the Add vertex icon in the floating window that appears, or use the A hotkey to activate ‘Add vertex mode’
3. Move the mouse, which should now be in a ’+’ icon, over the edge where you wish to add a vertex. The edge in question should highlight indicating you’ve placed the mouse correctly. Click to place a vertex
4. Continue placing vertices until you’re satisfied. Press A or click the Add vertex icon on the floating toolbar to turn off add vertex mode

Removing a polygon or polyline vertex - Label Editor V1

To remove vertices from a polygon::

1. Click on the polygon you want to remove a vertex from.
2. Click the Remove vertex icon in the toolbar that appears, or use the S hotkey to activate ‘Remove vertex mode’.
3. Move the cursor to the vertex you want to remove. The cursor should have a ‘cancel’ icon attached.
4. Click the vertex you want to remove.
5. Continue removing any vertices until you are satisfied.
6. Click the icon from step 2 again, or press the S hotkey again to exit ‘Remove vertex mode’.

Freehand edge touchup - Label Editor V1

If you need to modify the edge of a polygon to account for complicated shapes or other requirements, the freehand edge touchup feature allows you edit the edge of the polygon in a free-hand manner, from one vertex to another.

1. Click the polygon you would like to edit.
2. Click the Freehand edge touchup icon in the floating toolbar that appears, or use the F hotkey to activate ‘freehand edge touchup’ mode.
3. Mouse over the vertex you would like to start editing from, and click to start the edit process.
4. Drag the cursor freely to draw out the new polygon edge. Navigate the cursor to the pre-existing vertex you want to finish on, and click to finalize the new edge boundary.