When Apple released CreateML in 2018 it was a groundbreaking innovation. For the first time, developers could easily train a machine learning model with no code. No code, that is, except for the one-off scripts they needed to write to munge their data into the proper JSON format that was adopted from Apple's acquired machine learning framework, Turi Create.
Fortunately, Roboflow can both read and write CreateML JSON files so you can truly train a model and export it for use with an Apple iOS device via CoreML without writing a single line of code. Follow our CreateML tutorial to learn how to use it to train a no-code object detection model.
Below, learn the structure of CreateML JSON.
[
{
"image": "0001.jpg",
"annotations": [
{
"label": "helmet",
"coordinates": {
"x": 162.5,
"y": 45,
"width": 79,
"height": 88
}
},
{
"label": "person",
"coordinates": {
"x": 145.5,
"y": 176,
"width": 251,
"height": 350
}
}
]
}
]With Roboflow supervision, an open source Python package with utilities for completing computer vision tasks, you can merge and split detections in CreateML JSON. Read our dedicated guides to learn how to merge and split CreateML JSON detections.
Below, see model architectures that require data in the CreateML 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.