Coco utilities

Evaluation functions for FlatBug datasets.

flat_bug.coco_utils.annotations_2_contours(annotations: Dict[str, dict]) → Dict[str, List[array]]

Converts COCO annotations to contours.

Parameters:

annotations (Dict[str, dict]) – COCO annotations.

Returns:

Contours.

Return type:

out (Dict[str, List[np.array]])

flat_bug.coco_utils.annotations_to_numpy(annotations: List[Dict[str, int | List[int]]]) → Tuple[ndarray, ndarray]

Converts COCO annotations to NumPy arrays.

Parameters:

annotations (List[Dict[str, Union[int, List[int]]]]) – COCO annotations.

Returns:

Bounding boxes and contours.

Return type:

out (Tuple[np.ndarray, np.ndarray])

flat_bug.coco_utils.contour_area(c: ndarray) → ndarray

Calculates the area of a contour.

Parameters:

c (np.ndarray) – Contour of shape (n, 2).

Returns:

Scalar area of the contour of shape (1,).

Return type:

out (np.array[np.int32])

flat_bug.coco_utils.contour_bbox(c: ndarray) → ndarray

Calculates the bounding box of a contour.

Parameters:

c (np.ndarray) – Contour.

Returns:

Bounding box.

Return type:

oyut (np.ndarray)

flat_bug.coco_utils.fb_to_coco(d: Dict, coco: Dict) → Dict

Converts a FlatBug dataset to a COCO dataset.

Parameters:

  • d (dict) – FlatBug dataset.

  • coco (dict) – An instantiated COCO dataset or an empty dictionary.

Returns:

COCO dataset.

Return type:

out (dict)

flat_bug.coco_utils.filter_coco(coco: dict, confidence: float | None = None, area: int | None = None, verbose: bool = False) → dict

Filters COCO annotations by confidence.

Parameters:

  • coco (dict) – COCO dataset.

  • confidence (Optional[float], optional) – Confidence threshold. Defaults to None; no threshold.

  • area (Optional[int], optional) – Area threshold. Defaults to None; no threshold.

  • verbose (bool, optional) – Verbose mode. Defaults to False.

Returns:

Filtered COCO dataset.

Return type:

out (dict)

flat_bug.coco_utils.format_contour(c: List) → ndarray

Formats a contour to the OpenCV format.

Parameters:

c (list) – Contour.

Returns:

Formatted contour.

Return type:

out (np.ndarray)

flat_bug.coco_utils.split_annotations(coco: dict, strip_directories: bool = True) → Dict[str, dict]

Splits COCO annotations by image ID.

Parameters:

  • coco (dict) – COCO dataset.

  • strip_directories (bool, optional) – Flag to indicate whether only the basename of the images should be included in the result. Defaults to True.

Returns:

Dict of COCO datasets, split by image ID and keyed by image name.

Return type:

out (Dict[str, dict])