What are Annotations?
-
Print
-
DarkLight
The Dataloop platform uses annotations as a way to label or mark items.
Annotations
Annotations are a way to label specific sections or entire items.
Our platform has 9 different types of Annotations:
- Classification : Label entire items (except in audio and video )
- Point : Point at a small section (or use Pose for point of a pre-defined template)
- Bounding Box : Mark a section with a square
- Cuboid : Annotate 2d data on a 3d scale
- Polygon : Label a section by drawing straight lines that are connected in order to form a chain (Polyline refers to when a Polygon is left open)
- Ellipse : Mark a section with an ellipse shape
- Semantic Segmentation : Mark a section by drawing
- Notes Tool : Used as a direct feedback feature for Annotators and Reviewers in the QA flow.
- Item Description : Lets you add free text to your items as annotations
- Auto Annotation Models : Use outsource models (yolo/coco/voc) to automatically create annotations
JSON Code For Annotations
Each entity in our system has a JSON code that represents its data within our system.
It includes the entity's data related to our software and its metadata.
JSON is a minimal text-based data exchange format that is used primarily to transmit data between a server and web application
To learn more about JSON click here.
{
"annotations": [
{
"id": "5dbad837d0e58c1ee23651fb", (Annotation ID)
"datasetId": "5c90ded0cb6f61001c22592b", (Dataset ID)
"type": "box", (Annotation type)
"label": "License Plate", (Label/Class)
"attributes": [ (Attribute)
"Number visible"
],
"metadata": {
"system": {
"status": "approved", (QA status for annotation)
"startTime": 0, (This is the start time of the annotation)
"endTime": 1, (This is the end time of the annotation)
"frame": 0, (This is the first frame of the annotation)
"endFrame": 1, (This is the last frame of the annotation)
"snapshots_": [ (Snapshot information Relevant for video annotation)
{
"fixed": true, (Frame that was changed by a user if true or interpulated if false)
"type": "transition",
"frame": 0, (Snapshpt frame)
"objectVisible": true, (Status of annotation, true = visible and flase = hidden)
"data": [ (This are the coordination of the annotation in each snapshot)
{
"x": 99.32,
"y": 401.53,
"z": 0
},
{
"x": 236.35,
"y": 482.26,
"z": 0
}
],
"label": "License Plate", (Snapshot information Relevant for video annotation)
"attributes": [ (Snapshot information Relevant for video annotation)
"Number visible"
]
}
],
"parentId": "5dbad722a49562452ae28512", (The ID of parent annotation)
"automated": false, (Relevant for video annotation)
"objectId": "1234", (A Number assigned by a labeler for visual tracking)
"isOpen": false, (For Polygon annotation, False = Closed shape, True = Open shape)
"system": false, (System metadata)
"description": "A5G45J" (Annotation text description)
},
"user": {} (Optional metadata that can be added by user via SDK)
},
"creator": "user@dataloop.ai", (Annotation creator)
"createdAt": "2019-10-31T12:48:55.000Z", (Annotation creation date and time)
"updatedBy": "user@dataloop.ai", (Annotation edits by user name)
"updatedAt": "2019-10-31T12:49:02.000Z", (Annotation edits date and time)
"itemId": "5da6beb96376a7ab3b516e5b", (Image ID)
"url": "https://gate.dataloop.ai/api/v1/annotations/5dbad837d0e58c1ee23651fb", (API url for annotation)
"item": "https://gate.dataloop.ai/api/v1/items/5da6beb96376a7ab3b516e5b", (API url for image item)
"dataset": "https://gate.dataloop.ai/api/v1/datasets/5c90ded0cb6f61001c22592b", (API url for dataset)
"coordinates": [ (Annotation position coordinates)
{
"x": 99.32,
"y": 401.53,
"z": 0
},
{
"x": 236.35,
"y": 482.26,
"z": 0
}
]
}
],
"_id": "5da6beb96376a7ab3b516e5b", (Image ID)
"filename": "/folder_count/without/akona40.jpg" (Image path in the dataset)
}
Studio
You can annotate annotations using our studio.
When in the task assignment flow, reach the studio either by clicking on an assignment in the assignment page for annotators.
Or by clicking on an assignment in the assignment of tasks page for the rest of the roles.
Reach the assignment of tasks page by clicking on a specific task on the Tasks Page.
Outside of the task assignment flow, reach the studio by pressing an item in the dataset browser.
Enter the Dataset Browser by clicking on the dataset page or project dashboard, and clicking on the “BROWSE” icon of a dataset.
