coco
All format examples · Machine-readable schema · Convert outputs to COCO
Two images: one detection and one explicitly empty image. bbox is top-left x,y,width,height. Import returns two ISIR records, using image_id to associate annotations.
Example
Illustrative values, not predictions from a model. Unless stated otherwise, the image is 200 pixels wide and 100 pixels high.
{
"annotations": [
{
"area": 1600,
"bbox": [10, 20, 40, 40],
"category_id": 0,
"id": 10,
"image_id": 1,
"iscrowd": 0,
"score": 0.8
}
],
"categories": [
{
"id": 0,
"name": "insect"
}
],
"images": [
{
"file_name": "frame.jpg",
"height": 100,
"id": 1,
"width": 200
},
{
"file_name": "empty.jpg",
"height": 100,
"id": 2,
"width": 200
}
]
}Convert to a standard format
Use Converter.convert with source="coco" and target="coco". Your application supplies the predictions and any required image context or producer settings. The adapter validates whether the output contains the information the target needs.
This importer accepts multiple images: set cardinality="collection".
General conversion example and required context
Inspect the intermediate representation for this example
This developer example prints ISIR, the internal representation used between adapters. Run from the repository checkout; no model download is needed.
python3 examples/formats/to_isir.py cocoHow to interpret it
Scope: COCO dataset annotation structure; prediction-only COCO result lists are a different interface.
Bbox: One [x, y, width, height] tuple per annotation, in pixels; x,y identify the top-left corner.
Metadata: Dataset info/licenses and image license/URL/date metadata may be absent. The retained Flatbug 1.2.0 COCO probe contains an empty year and date_captured=0; these placeholders are preserved, not invented timestamps.