vis4d.eval.bdd100k

BDD100K evaluators.

class BDD100KDetectEvaluator(annotation_path, config_path, mask_threshold=0.0)

BDD100K 2D detection evaluation class.

__init__(annotation_path, config_path, mask_threshold=0.0)

Initialize the evaluator.

__repr__()

Concise representation of the dataset evaluator.

Return type:

str

class BDD100KSegEvaluator(annotation_path)

BDD100K segmentation evaluation class.

__init__(annotation_path)

Initialize the evaluator.

__repr__()

Concise representation of the dataset evaluator.

Return type:

str

property metrics: list[str]

Supported metrics.

gather(gather_func)

Gather variables in case of distributed setting (if needed).

Parameters:

gather_func (Callable[[Any], Any]) – Gather function.

Return type:

None

reset()

Reset the evaluator.

Return type:

None

process_batch(data_names, masks_list)

Process tracking results.

Return type:

None

evaluate(metric)

Evaluate the dataset.

Return type:

tuple[Dict[str, Union[float, int, Tensor]], str]

class BDD100KTrackEvaluator(annotation_path, config_path='box_track', mask_threshold=0.0)

BDD100K 2D tracking evaluation class.

__init__(annotation_path, config_path='box_track', mask_threshold=0.0)

Initialize the evaluator.

__repr__()

Concise representation of the dataset evaluator.

Return type:

str

property metrics: list[str]

Supported metrics.

evaluate(metric)

Evaluate the dataset.

Return type:

tuple[Dict[str, Union[float, int, Tensor]], str]

Modules

detect	BDD100K detection evaluator.
seg	BDD100K segmentation evaluator.
track	BDD100K tracking evaluator.