The Visual Geometry Group (VGG) at the University of Oxford released an open source annotation tool called VIA (VGG Image Annotator). It outputs its labels to CSV or JSON; Roboflow supports importing both and converting them to any other object detection format you like.
See our VGG Image Annotator Tutorial to learn how to use VIA to create object detection annotations.
Below, learn the structure of VGG Image Annotator JSON.
{
"img0001": {
"filename": "img0001.png",
"size": 2512968,
"regions": [{
"shape_attributes": {
"name": "rect",
"x": 827,
"y": 890,
"width": 150,
"height": 651
},
"region_attributes": {
"type": "helmet"
}
}, {
"shape_attributes": {
"name": "rect",
"x": 1943,
"y": 875,
"width": 120,
"height": 639
},
"region_attributes": {
"type": "head"
}
}],
"file_attributes": {}
}
}With Roboflow supervision, an open source Python package with utilities for completing computer vision tasks, you can merge and split detections in VGG Image Annotator JSON. Read our dedicated guides to learn how to merge and split VGG Image Annotator JSON detections.
Below, see model architectures that require data in the VGG Image Annotator 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.