Converters

Import formats

You can import multiple annotation formats to SuperAnnotate with the supported conversion functions.

To convert the annotation format from COCO instance segmentation to SuperAnnotate for an Image Project:

sa.import_annotation(
    input_dir = "./coco", 
    output_dir = "./sa_vector", 
    dataset_format = "COCO", 
    dataset_name = 'instances_train2017', 
    project_type = 'Vector', 
    task='instance_segmentation')

To convert the annotation format from COCO instance segmentation to SuperAnnotate for an Image (Legacy) Project:

sa.import_annotation(
    input_dir = "./coco", 
    output_dir = "./sa_pixel", 
    dataset_format = "COCO", 
    dataset_name = 'instances_train2017', 
    project_type = 'Pixel', 
    task='panoptic_segmentation')

SuperAnnotate supports conversions from common annotation formats like COCO, Pascal VOC, YOLO, and more.

Check the table below for the list of supported formats and task combinations in the import_annotation function.

Format	Project Type	Task
COCO	Image	`instance_segmentation`
COCO	Image	`object_detection`
COCO	Image (Legacy)	`panoptic_segmentation`
VOC	Image	`instance_segmentation`
VOC	Image	`object_detection`
YOLO	Image	`object_detection`

Export Formats

SuperAnnotate's export format can be converted to COCO:

sa.export_annotation(
    input_dir = "./exports/sa_vector", 
    output_dir = "./coco_panoptic", 
    dataset_format = "COCO", 
    dataset_name = "panoptic_train2017", 
    project_type='Pixel', 
    task='panoptic_segmentation')

The export converter supports the following COCO tasks:

Format	Project Type	Task
COCO	Image (Legacy)	`panoptic_segmentation`
COCO	Image (Legacy)	`instance_segmentation`
COCO	Image	`instance_segmentation`
COCO	Image	`object_detection`

You can also convert annotations from the SuperAnnotate format into Pandas DataFrame where each annotated instance is stored as a separate row:

annotations_df = sa.aggregate_annotations_as_df(
        project_root = “./exports/video_annotations”, 
        project_type = “Video”, 
        folder_names = [“test”, “train”])

You can further filter the obtained DataFrame by custom Pandas queries.

The columns of the returned dataframe will vary depending on the project type.

Image projects

Column name	Column description
imageName	Image name
folderName	Folder name of the annotations JSON
imageHeight	Image height
imageWidth	Image width
imageStatus	Image’s annotation status
imagePinned	Image’s pin status
imageAnnotator	ID of the assigned annotator
imageQA	ID of the assigned annotator QA
tag	Video tag
commentResolved	Information whether the comment was resolved or not
instanceId	The instance’s index in the image’s instances list
className	Class name
classColor	Class color
attributeGroupName	Attribute group name
attributeName	Attribute name
type	Instance type
error	Instance’s approval status
locked	Instance locked status
visible	Instance’s visibility
trackingId	Instance’s tracking ID
probability	Instance’s confidence score
pointLabels	Object that contains the instance’s point labels
meta	Timestamp’s meta information. Example: bounding box’s points
groupId	Instance’s group ID
createdAt	Instance’s creation time
creationType	Instance’s creation type
creatorRole	Instance’s creator role
creatorEmail	Instance’s creator email
updatedAt	The last time the instance was updated
updatedBy	The ID of the user who last updated the instance
updatorRole	The role of the user who last updated the instance

Video projects

Column name	Column description
videoName	Video name
folderName	Folder name of the annotation JSON
videoHeight	Video’s height
videoWidth	Video’s width
videoStatus	Video’s annotation status
videoURL	Video’s source URL
videoDuration	Video duration
videoError	Video approval
videoAnnotator	The ID of the assigned Annotator
videoQA	The ID of the assigned QA
tagId	The tag’s index in the video’s tags list
tag	Tag name
instanceId	The instance’s index in the video’s instances list
instanceStart	The instance’s start time (in microseconds)
instanceEnd	The instance’s end time (in microseconds)
type	Instance’s type
className	Instance’s class name
createdAt	Instance’s creation time
createdBy	The ID of the user who created the instance
creatorRole	The role of the user who created the instance
updatedAt	The time the instance was last updated
updatedBy	The ID of the user who updated the instance
updatorRole	The role of the user who last updated the instance
pointLabels	Object that contains the instance’s point labels
parameterId	The index of the parameter in the parameters list of the instance
parameterStart	Parameter’s start time
parameterEnd	Parameter’s end time
timestampId	The timestamp’s index in the list of timestamps
meta	Timestamp’s meta information. Example: bounding box’s points
attributeId	The index of the attribute in the attributes list of the timestamp
attributeGroupName	Attribute’s group name
attributeName	Attribute’s name