SuperAnnotate Docs

Explore a dataset

Generate a dataset of a Vector project’s images to check the quality of the annotations in a grid view.

Generate a dataset

  1. Go to the Images tab.
  2. Select the Explore button.
  3. In the popup, choose Explore.

When the dataset generation completes successfully, select the arrow icon to go to the Explore page. The Explore page features all the project’s images.

A dataset can have three statuses: In progress, Complete, and Failed.

📘

Only Team owners and Team admins can generate and explore datasets.

❗️

  1. You cannot delete a dataset.
  2. You cannot generate more than one dataset. Selecting the Explore button after generating a dataset will overwrite the current dataset.
  3. You need at least 8 images in the project to be able to generate a dataset.

Explore instances

The Explore page gives you a grid view of all the image search results. You can see the instances of all the images as well as the following information:

  • Image name
  • Image status
  • Copy button to copy the image name
  • QA name
  • Annotator name
  • Instances
  • Classes
  • Tags
  • Attributes

You can show and hide these items in Display settings.

Select the copy and paste icon to copy the image name and click on an image to open it in the editor.

Search for instances

To search for instances, insert a query in the input bar and press Enter to run the query. The results displayed are images that contain instances that match the query. For example, if you searched for instances of pedestrians annotated with a bounding box, the search results are images that contain pedestrians annotated with a bounding box.

Query format

Single parameter
main field.field.subfield=parameter 1

Example
[email protected]

Array of parameters
main field.field.subfield IN [parameter 1, parameter 2…]

Example
metadata.lastAction.email IN [[email protected], user[email protected]]

Array of parameters and more than one field
main field.field.subfield IN [parameter 1, parameter 2…] OR main field.field.subfield IN [parameter 1, parameter 2…]

Example
metadata.lastAction.email IN [[email protected], [email protected]] OR comments.resolved=true

Operators

OperatorMeaning
=Equal to
!=Not equal to
>Greater than
<Less than
>=Greater than or equal to
<=Less than or equal to
()Shows the hierarchy of queries
ANDReturns True if both statements are true
ORReturns True if either statements are true
INSet of values to consider

Fields

The field names and values should be written in the JSON format.

Instances

FieldValue
instances.typebbox, polygon, point, ellipse, cuboid, polyline, or rbbox.
instances.class.nameString. If there is no class assigned to an instance, then the parameter should be none.
instances.probabilityDouble
instances.pointLabelsDouble
attributes.nameString
attributes.groupNameString
errorIf the instance is approved, then it is false. If it is disapproved, then it is true. If there is no approval, then it is null. If the instance is neither approved nor disapproved, then the error field will not appear in JSON.
createdAtDate in this format: YYYY-MM-DD.
createdBy.emailEmail address
createdBy.roleCurrent SuperAnnotate role
creationTypemanual or pre-annotation
updatedAtDate in this format: YYYY-MM-DD.
updatedBy.emailEmail address
updatedBy.roleCurrent SuperAnnotate role
countNumber of instances in the image

Metadata

FieldValue
lastAction.emailEmail address
widthWidth in pixels
heightHeight in pixels
nameString
isPredictedBoolean: true or false.
annotatorEmailEmail address
qaEmailEmail address
statusImage status
pinnedBoolean: true or false.

Comments

FieldValue
resolvedBoolean: true or false.
countNumber of comments

Tags

FieldValue
TagsTag name

Folder

FieldValue
folder.nameFolder name in a string format

Download dataset

To download a JSON file of your dataset:

  1. Go to the Explore page.
  2. Run a query (optional).
  3. Select Download.

🚧

If you do not run a query, the whole dataset will be downloaded.

Learn more about JSON files for Vector projects.

If you choose to include annotations and statuses when copying an image, the following fields will be excluded from the JSON file:

  • “pinned”: false
  • “annotatorEmail”: null
  • “qaEmail”: null

🚧

You cannot download an expired dataset. If your dataset is expired, go back to the Images tab, select Explore to regenerate the dataset, and try downloading again.

Display settings

Adjust the settings of the Explore page in Display settings.

Column view

Choose how many images you want to show per column. The minimum number of images per column is one image and the maximum is five images.

Opacity

Drag the slider to the left or to the right to decrease or increase the opacity of the instances.

Image labels

Choose what items you want to show or hide on the images by checking or unchecking the corresponding boxes. The items are:

  • Information (Image name, QA name, Annotator name, and annotation status)
  • Tags
  • Classes
  • Attributes

To choose which attributes you want to show or hide on the images, select the arrow next to Attributes and check (show) or uncheck (hide) the boxes.

Instances

Choose which instance(s) you want to show on the images by selecting the corresponding boxes. If you want to show all the instances, check the All instances box. Use the search bar to find instances.

📘

Select Reset to apply default settings.

Permissions

As a Team owner, you can give a Team admin permission to generate a dataset.

  1. Go to My teams.
  2. Select Permissions.
  3. Check the Allow to explore annotations box.

Updated 12 days ago


Explore a dataset


Suggested Edits are limited on API Reference Pages

You can only suggest edits to Markdown body content, but not to the API spec.