Datasets

class flat_bug.datasets.FlatBugYOLODataset(max_instances: int | None, classes: None = None, subset_args: Dict | None = None, *args, **kwargs)

build_transforms(hyp: IterableSimpleNamespace) → Compose

Builds and appends transforms to the list.

Parameters:

hyp (dict, optional) – Hyperparameters for transforms.

Returns:

Composed transforms.

Return type:

(Compose)

cache_labels(path: Path = PosixPath('labels.cache'))

OBS: DO NOT USE THIS FUNCTION MANUALLY.

load_image(i: int | slice) → Tuple[ndarray, Tuple[int, int], Tuple[int, int]]

Load an image from dataset index ‘i’.

Parameters:

• i (int) – Index of the image to load.

• rect_mode (bool, optional) – Whether to use rectangular resizing.

Returns:

Loaded image as a NumPy array. (Tuple[int, int]): Original image dimensions in (height, width) format. (Tuple[int, int]): Resized image dimensions in (height, width) format.

Return type:

(np.ndarray)

Raises:

FileNotFoundError – If the image file is not found.

class flat_bug.datasets.FlatBugYOLOValidationDataset(max_instances: int | None, classes: None = None, subset_args: Dict | None = None, *args, **kwargs)

build_transforms(hyp: IterableSimpleNamespace) → Compose

Builds and appends transforms to the list.

Parameters:

hyp (dict, optional) – Hyperparameters for transforms.

Returns:

Composed transforms.

Return type:

(Compose)

flat_bug.datasets.calculate_image_weights(image_paths: List[str]) → List[float]

Calculate normalized weights for each image based on the file sizes, normalized by the minimum file size, so that the values are between 1 and infinity.

Parameters:

image_paths (List[str]) – List of image file paths.

Returns:

normalized weights for each image.

Return type:

out (List[float])

flat_bug.datasets.generate_indices(weights: List[float], target_size: int | None = None) → List[int]

Deterministically generates a list of indices based on the provided weights to oversample the items.

Parameters:

• weights (List[float]) – List of weights for each item.

• target_size (Optional[int], optional) – Desired size of the output list. If None, the size of the output is approximately the sum of the weights.

Returns:

List of indices to oversample the items.

Return type:

out (List[int])

flat_bug.datasets.reweight(weights: List[float], target_sum: float | int) → List[float]

Reweights the provided list of weights so that their sum equals the target sum.

Parameters:

• weights (List[float]) – List of weights to reweight.

• target_sum (Union[float, int]) – Desired sum of the weights.

Returns:

Reweighted weights.

Return type:

out (List[float])

flat_bug.datasets.subset(self: FlatBugYOLODataset, n: int | None = None, pattern: str | None = None)

Subsets the dataset to the first ‘n’ elements that match the pattern.

Parameters:

• n (Optional[int], optional) – The number of elements to keep. Defaults to None; keep all.

• pattern (Optional[str], optional) – A regex pattern to match the filenames. Defaults to None; match all.