Once you're ready to get serious about a computer vision project, you are probably going to need to outsource your labeling. One popular service for this is Scale.com. Unfortunately, the results they send back are not in a format that is ready to go into common machine learning models.
That's where Roboflow comes in. Simply drop in the JSON files you receive from Scale's human annotators and we'll convert them to whatever format you need for your model.
Below, learn the structure of Scale AI JSON.
[
{
"task_id": "a9b7c5d3e1f",
"created_at": "2000-01-01T00:00:00.000Z",
"completed_at": "2000-01-01T00:00:00.000Z",
"callback_url": "example@example.com",
"type": "annotation",
"status": "completed",
"instruction": "\n# Instructions\n\nLabel the helmets on the workers.",
"params": {
"attachment": "https://storage.googleapis.com/roboflow/0001.jpg",
"attachment_type": "image",
"objects_to_annotate": [
"helmet",
"head",
"person"
],
"with_labels": true,
"min_width": 0,
"min_height": 0,
"examples": []
},
"is_test": false,
"urgency": "standard",
"metadata": {},
"callback_succeeded": true,
"processed_attachments": [],
"project": "Hard Hat Workers",
"response": {
"annotations": [
{
"width": 189,
"height": 144,
"label": "helmet",
"left": 1374,
"top": 2128
},
{
"width": 157,
"height": 122,
"label": "person",
"left": 1427,
"top": 2409
}
]
},
"project_param_version": 0,
"scale_internal_attachment": "https://d2qt2k6tuv6mdh.cloudfront.net/a9b7c5d3e1f/001.jpg",
"workStarted": true
}
]With Roboflow supervision, an open source Python package with utilities for completing computer vision tasks, you can merge and split detections in Scale AI JSON. Read our dedicated guides to learn how to merge and split Scale AI JSON detections.
Below, see model architectures that require data in the Scale AI JSON format when training a new model.
On each page below, you can find links to our guides that show how to plot predictions from the model, and complete other common tasks like detecting small objects with the model.