fiftyone.utils.data.parsers.ImageDetectionSampleParser

Generic parser for image detection samples whose labels are represented as fiftyone.core.labels.Detections instances.

This implementation supports samples that are (image_or_path, detections_or_path) tuples, where:

• image_or_path is either an image that can be converted to numpy format via np.asarray() or the path to an image on disk

• detections_or_path can be any of the following:

  • None, for unlabeled images

  • a list of detections in the following format:

    [
        {
            "<label_field>": <label-or-target>,
            "<bounding_box_field>": [
                <top-left-x>, <top-left-y>, <width>, <height>
            ],
            "<confidence_field>": <optional-confidence>,
            "<attributes_field>": {
                <optional-name>: <optional-value>,
                ...
            }
        },
        ...
    ]
    

    In the above, label-or-target is either a class ID (if classes is provided) or a label string, and the bounding box coordinates can either be relative coordinates in [0, 1] (if normalized == True) or absolute pixels coordinates (if normalized == False). The confidence and attributes fields are optional for each sample.

    The input field names can be configured as necessary when instantiating the parser.

  • the path on disk to a file in the above format

  • a fiftyone.core.labels.Detections instance