Annotations
Get annotations into memory
To get image annotations into the memory:
annotations = sa.get_annotations(
project = "Project Name",
items = ["item_name.png"])
This will return a Python dictionary with a annotation_json
key with the value of the annotations in the SuperAnnotate format.
import json
print(json.dumps(annotations["annotation_json"], indent = 4, sort_keys = True))
This is the output:
{
"instances": [
{
"attributes": [],
"classId": 816825,
"className": "apple",
"createdAt": "2021-06-09T11:24:54.794Z",
"createdBy": {
"email": "[email protected]",
"role": "Annotator"
},
"creationType": "Manual",
"error": null,
"groupId": 0,
"locked": false,
"pointLabels": {},
"points": {
"x1": 126.35,
"x2": 384.42,
"y1": 748.26,
"y2": 921.96
},
"probability": 100,
"trackingId": null,
"type": "bbox",
"updatedAt": "2021-07-27T14:21:52.904Z",
"updatedBy": {
"email": "[email protected]",
"role": "QA"
},
"visible": true
}
],
"metadata": {
"height": 1024,
"width": 1024,
"name": "fruit.png",
"lastAction": {
"email": "[email protected]",
"timestamp": 1627395705911
}
},
"tags": [],
"comments": []
}
Download annotations
To download image annotations:
sa.download_image_annotations(
project = "Project Name",
image_name = "image_name.png",
local_dir_path = "./annotations")
Use the
prepare_export
functionality to download images or annotations in bulk. Learn more here.
To download annotations from a project's root:
sa.download_annotations(
project="Project 1",
path="/Users/projectadmin/Desktop/Annotations"
)
To download all the annotations from a project:
sa.download_annotations(
project="Project 1",
path="/Users/projectadmin/Desktop/Annotations",
recursive=True
)
To download specific items:
# step 1: query completed annotations and store item names
completed_item_names = [
item['name'] for item in sa.query(
project="Image Project",
query="metadata(status = Completed)"
)
]
print(f'there are {len(completed_item_names)} completed items names are: {completed_item_names}')
# step 2: download annotations by providing the list of item names
sa.download_annotations(
project="Image Project",
path="/Users/projectadmin/Desktop/Annotations",
items = completed_item_names,
recursive=True
)
To download annotations using a specific JSON structure:
# step 1: define the converter function that accepts the SA annotation dict as an argument
def convert(sa_annotation):
filtered_boxes = [
instance for instance in sa_annotation['instances'] if instance['type'] == 'bbox'
]
boxes = list(map(lambda box:
{
'type': box['type'],
'class': box['className'],
'coordinates': box['points']
}, filtered_boxes)
)
return {'objects': boxes}
#step 2: pass the converter function to callback parameter
sa.download_annotations(
project="Project 1",
path="/Users/projectadmin/Desktop/Annotations",
items=["Image 1.jpeg"],
recursive=True,
callback=convert
)
Updated 10 months ago