Explore dataset (Image Projects)

Go to Explore to search images in Image Projects using more specific criteria than just image names. The fields and values entered in the query bar match the JSON format.

You can view all the images in a list view or a grid view. The grid view is the default view. To switch to the grid view, click the grid button next to the query bar.

The grid view allows you to see your images and their annotations in greater detail. It also shows you the names of the assigned Annotators and QAs, the image name and status, and the approval status.

The instance view allows you to see each instance separately. You can also filter, approve/disapprove, and delete instances which will, in turn, help you manage your dataset’s health in a more effective way.

How does it work?

Explore works in two ways: you can type the queries yourself or use autosuggestions. Autosuggestion makes it easier for you to generate the queries.

To generate a query:

  1. Go to the Explore tab.
  2. If the data isn't synced, click Sync data. If the dataset is synced, skip to Step 3.
  3. Click the input bar.
  4. Choose the suggested fields and operators or type the query yourself.
  5. To run the query, press Enter on the keyboard.

📘

  • If your permission to use Explore was disabled and then enabled, then you’ll have to sync the data again.
  • If you're experiencing issues with the dataset, you can click Resync the dataset.

Grid view

Column view

Choose how many items there are per row. The number of items per row is 3 by default.

Display settings

In Display Settings, you can modify what you see in the Explore tab (grid view only).

Opacity

Modify the opacity of the instances by dragging the slider to the left or right.

Item information

  • Check the Information box to show the approval status, item name, and item status.
  • Check the Classes box to show the names of the classes.
  • Check the Attributes box to show the names of the attributes.

Class name

Check/Uncheck the box of a particular class to show/hide instances with that class.

Approval status

Show or hide approved and disapproved instances. You can also show or hide instances that don’t have an approval status.

Instance type

Show or hide certain instance types. You can see all the tags at the bottom of each item. To hide the tags, go to Instance types and uncheck the Tag box. Note that the attribute groups and attributes are also shown when you hover over a tag.

Group ID

Type a group ID to see all the instances that have that group ID.

Probability

Show or hide your instances by their probability score.

Creation type

Show or hide your instances by their creation type:

  • Manual (the instance was created manually in the editor)
  • Prediction (the instance was created with a model)
  • Pre-annotation (the instance was imported)

Created at

Show or hide your instances by creation date.

Updated at

Show or hide your instance by the date they were updated.

Instance view

Display

Columns

Choose how many items there are per row. The number of items per row is 3 by default.

Display settings

In Display Settings, you can modify the view of the items.

Opacity

Modify the opacity of the instances by dragging the slider to the left or right.

Item information
  • Check the Information box to show the approval status and item name.
  • Check the Classes box to show the names of the classes.
  • Check the Attributes box to show the names of the attributes.

Filter

Class and attribute

Check/Uncheck the box of a particular class to show/hide instances with that class and attributes.

Approval status

Filter approved or disapproved instances. You can also show or hide instances that don’t have an approval status.

Instance type

Filter by instance types: Polyline, Polygon, Bounding Box, Rotated Bounding Box, Ellipse, Point, and Cuboid,

Group ID

Type a group ID to see all the instances that have that group ID.

Probability

Filter your instances by their probability score.

Creation type

Filter your instances by their creation type:

  • Manual (the instance was created manually in the editor)
  • Prediction (the instance was created with a model)
  • Pre-annotation (the instance was imported)

Created at

Filter your instances by creation date.

Updated at

Filter your instances by the date they were updated.

Manage instances

Approve/Disapprove instances

To approve/disapprove an instance or multiple instances:

  1. Select an instance or multiple instances.
  2. Click Approve/Disapprove instances.
  3. Choose the approval status: Approve, Disapprove, or None.
  4. In the popup, click Change.

📘

Messages

  • When you change the approval status successfully, you’ll see this message: You've successfully changed the approval status of [number of instances] instance(s).
  • When it took a while to change the approval status, you’ll see this message: [number of updated instances]/[total number of instances] instances updated.
  • If there’s an error, you’ll see this message: Oops! Something went wrong.

Delete instances

To delete an instance or multiple instances:

  1. Select an instance or multiple instances.
  2. Click Delete.
  3. In the popup, click Delete.

📘

Messages

  • When the items are deleted successfully, you’ll see this message: You've successfully deleted [number of instances] instance(s).
  • When the items are taking too long to be deleted, you’ll see this message: [number of updated instances]/[total number of instances] instances deleted
  • When the items couldn’t be deleted, you’ll see this message: Oops! Something went wrong.

Example queries

FunctionQuery
To search for items that have the In Progress status and are have Smart Prediction appliedmetadata(isPredicted = true AND status = InProgress )
To search for items that have instances that are polygons and are assigned to the classes car or pedestrianinstance(type = polygon AND className IN [Car, Pedestrian] )
To search for items that have unresolved commentscomment(resolved = false )
To search for items that contain Bounding Boxes instance(type = bbox )
To search for items that contain instances with the class Carinstance(className = Car )
To search for items that have the Completed status and are located in folder1metadata(status = Completed ) AND folder = folder1
To search for items that contain more than 3 instances with the class Car and the attribute BlackinstanceCount(className = Car AND attributes.name = Black ) > 3
To search for pinned itemsmetadata(pinned = true )

📘

Click on the query bar to see the last 3 unique queries.

Fields and values

Instance and instance count

You can search instances (instance) and instance counts (instanceCount) by the following fields:

FieldDescriptionValue
typeInstance typebbox polygon polyline point ellipse, cuboid rbbox tag
classNameClass nameThe value is a class name.
probabilityModel training confidence scoreNumeric value
attribute.nameAttribute nameThe value is an attribute name.
attribute.groupNameAttribute group nameThe value is an attribute group name.
errorInstance’s approval statustrue false null
createdBy.emailEmail address of the project contributor who created the instance.The value is an email address.
createdBy.roleRole of the project contributor who created the instance.Annotator QA Admin
creationTypeThe instance’s creation type.Manual Preannotation Prediction
updatedBy.emailEmail address of the project contributor who made the last change to the instance.The value is an email address.
updatedBy.roleRole of the project contributor who made the last change to the instance.Annotator QA Admin
createdAtDate when the instance was created.The value is a date with this format YYYY-MM-DD.
updatedAtDate when the instance was last updated.The value is a date with this format YYYY-MM-DD.

🚧

About class names

If the class name contains an operator (and, or, in, not in, contains), the query will handle them as operators, not as words in class names. For example, if the class name is Cars and pedestrians, the query will handle the word cars and the word pedestrians separately. Instead, you can use symbols, such as underscores, to separate the words. So Cars and pedestrians would be changed to cars_and_pedestrians.

Metadata

You can search the metadata (metadata) by the following fields:

FieldDescriptionValue
lastAction.emailEmail address of the project contributor who made the last change to the item.The value is an email address.
lastAction.timestampDate when the item was last updated.The value is a date with this format YYYY-MM-DD.
widthItem widthNumeric value
heightItem heightNumeric value
nameItem file nameThe value is a name.
isPredictedInformation whether Smart Prediction was used on the item or not.true false
annotatorEmailThe assigned annotator’s email address.The value is an email address.
qaEmailThe assigned QA’s email address.The value is an email address.
statusItem statusNotStarted InProgress QualityCheck Returned Completed Skipped
pinnedItem’s pin statustrue false
approvalStatusItem’s approval statustrue false null
priorityPriority scoreNumeric value

Comment

You can search comments (comment) by the following field:

FieldDescriptionValue
resolvedInformation where the comment is resolved or not.true false

Comment count

You can search the comment count (commentCount) by the following field:

FieldDescriptionValue
commentCountThe number of comments the item has.Numeric value.

Tag

You can search tags (tag) this way:

tag [insert operator here] [insert tag name here]

Example: tag = nature

Folder

You can search folders by the following fields:

FieldDescriptionValue
folderNameFolder nameThe value is a name.
folderStatusFolder statusNotStarted InProgress Completed OnHold
Undefined

To search for root folders:

folder [insert operator here] root

Example: folder = root

📘

If the status of a folder has been changed, refresh the Explore tab for the queries to work correctly.

Random

To get a randomized subset of data:

OperatorDescriptionValue
randomUsed to get a randomized subset of dataNumeric value.

When you run a query with random, you’ll get a randomized subset of data. This will help you increase efficiency in checking the quality of the annotations and the dataset’s health.

So, if the query is metadata(status = InProgress ) random 50, 50 random items from the project that have the in In Progress status will be returned.

Custom metadata

You can search by custom metadata (customMetadata) this way:

FieldDescriptionValue
customMetadataCustom metadataDepending on the field type. Please check this page for more information.

Operators

OperatorDescription
=Equal
!=Not equal
>Greater than
<Less than
>=Greater than or equal to
<=Less than or equal to
()Query parentheses
[]Array
ANDReturns True if both statements are true
ORReturns True if either statements are true
INSet of values to consider
NOT INSet of values to not consider
CONTAINSUsed to search for a substring and returns the results containing that substring

📘

When the attribute has a numeric type, the following operators will be supported for attributes.name: > < =< !=

Sort list

You can sort the list of items in Explore by:

FieldDescriptionValue
metadata.nameItem nameASC (ascending) DESC (descending)
metadata.statusItem statusASC (ascending) DESC (descending)
metadata.lastAction.dateDate when the item was last updated.ASC (ascending) DESC (descending)
metadata.qaEmailEmail address of the assigned QA.ASC (ascending) DESC (descending)
metadata.annotatorEmailEmail address of the assigned annotator.ASC (ascending) DESC (descending)
commentCountThe number of comments the item has.ASC (ascending) DESC (descending)
instanceCountThe number of instances the item has.ASC (ascending) DESC (descending)

Save query

To save a query:

  1. Type a query in the query bar.
  2. Click on the star at the right edge of the query bar.
  3. In the popup, type a query name.
  4. Click Save.

📘

  • The maximum number of characters for a query is 60.
  • You can save up to 50 queries.
  • If you’ve successfully saved the query, you'll see this message: “You’ve successfully saved the query.”
  • If the query can’t be saved, you'll see this message: “Oops! Something went wrong.”

Rename query

To rename a query:

  1. Click on the star in the query bar.
  2. Click the three dots next to the query you want to rename.
  3. Select Rename.
  4. Type a new query name.
  5. Click Rename.

📘

If you’ve successfully renamed the query, you'll see this message: “You’ve successfully renamed the query.”
If the query can’t be renamed, you'll see this message: “Oops! Something went wrong.”

Delete query

To delete a query:

  1. Click on the star in the query bar.
  2. Click the three dots next to the query you want to rename.
  3. Select Delete.

📘

  • If you’ve successfully deleted the query, you'll see this message: “You’ve successfully deleted the query.”
  • If the query can’t be deleted, you'll see this message: “Oops! Something went wrong.”

Subsets

Add items to a subset

To add an item or multiple items to a subset:

  1. In Explore, select an item or multiple items.
  2. Click Add to subset.
  3. Choose an existing subset or create a new one. To create a new subset, type a non-existing name in the search bar and click Create subset.
  4. Click Save.

Copy items to a subset

To copy an item or multiple items from one subset to another:

  1. Go to the subset.
  2. Select an item or multiple items.
  3. Click the Copy/Move items button.
  4. Choose the subset where you want to copy the item(s).
  5. Click Copy.

📘

Messages

  • If you’ve successfully copied the items, you’ll see this message: You added [number of items] item(s) to the selected subset(s).
  • If the items are taking too long to be copied, you’ll see this message: You added [number of added items]/[number of total items] item(s) to the selected subset(s).
  • If something goes wrong, you’ll see this message: Oops! Something went wrong.

Move items from one subset to another

To move an item or multiple items from one subset to another:

  1. Go to the subset.
  2. Select an item or multiple items.
  3. Click the Copy/Move items button.
  4. Go to the Move tab.
  5. Choose the subset where you want to copy the item(s).
  6. Click Move.

📘

Messages

  • If you’ve successfully moved the items, you’ll see this message: You successfully moved [number of items] item(s).
  • If the items are taking too long to be moved, you’ll see this message: [number of updated items]/[number of total items] item(s) moved
  • If something goes wrong, you’ll see this message: Oops! Something went wrong.

View subset

To view all your subsets, click on Display Settings next to the search bar. You’ll find all your subsets under Subsets.

Rename subset

To rename a subset:

  1. Find the subset you want to rename under Subsets in the left panel.
  2. Click on the three dots > Rename.
  3. In the popup, type the new subset name.
  4. Click Rename.

📘

Messages

When you successfully rename a subset: “You’ve successfully renamed the subset.”
When something goes wrong while renaming a subset: “Oops! Something went wrong.”
In case the subset name that was entered already exists: “The subset couldn’t be renamed because the name already exists.”

Delete a subset

To delete a subset:

  1. Find the subset you want to delete under Subsets in the left panel.
  2. Click on the three dots > Delete.
  3. In the popup, select Delete.

Delete items from a subset

To delete items from a subset:

  1. Select items in a subset.
  2. Click Remove From Subset.

📘

You can perform the following bulk actions in a subset: change the status of items, approve/disapprove items, set priority scores, and download a dataset.

Change status in bulk

You can change the status of the item(s) that appear as a result of your query.

To change the status of an item or multiple items:

  1. Run a query.
  2. Select an item or multiple items.
  3. Click Change Status.
  4. Choose a status.
  5. Click Change.

Approve/Disapprove items in bulk

To change the approval status of an item or multiple items:

  1. Select one item or multiple items.
  2. Click the Approve/Disapprove items button in the menu at the bottom.
  3. Choose the status: Approve, Disapprove, or None.
  4. Click Change.

Set priority scores in bulk

To set the priority scores in bulk:

  1. Select multiple items.
  2. Click Priority scores.
  3. Choose a value.
  4. Click Set.

🚧

Annotators will receive the items with a higher priority score first.

Download items

To download items from the selected items in the results:

  1. Go to the Explore tab.
  2. Select the items you want to download.
  3. Click Download.
  4. When the download is ready, you’ll be redirected to the Download page. Find your dataset and click the download button to download it.

If you want to send the dataset to someone, click the link button and copy the link. Anyone who has the link can access the data.

📘

Download limitations

A user can generate up to 50 downloads per day and per project. Only up to 50 downloads can be available. If you have 50 downloads in a project and create a new one, the oldest download will be deleted to make space for the new project.

The download package contains:

  • A folder that contains the classes in JSON files.
  • A folder that contains the annotations in JSON files.

If some of the items are inside folders, then their corresponding files will be located inside folders in the download package as well. The folder structure is kept the same.