gunz_cm.resolution_enhancements.datasets package

Submodules

gunz_cm.resolution_enhancements.datasets.csr module

gunz_cm.resolution_enhancements.datasets.memmap_v1 module

Module.

Examples

class gunz_cm.resolution_enhancements.datasets.memmap_v1.MemmapDatasetV1(memmap_fpath: str, win_size: int, stride_size: int = None, transformations: Callable | None = None)

Bases: Dataset

A PyTorch Dataset class for loading patches from a memory-mapped file.

Args:

memmap_fpath (str): The path to the memory-mapped file. win_size (int): The size of the window/patch to extract from the memory-mapped file. stride_size (int, optional): The stride size to use when extracting patches. Defaults to win_size. transformations (callable, optional): A function/transform to apply to the data. Defaults to None.

Attributes:

_f (numpy.memmap): The memory-mapped file. win_size (int): The size of the window/patch to extract from the memory-mapped file. stride_size (int): The stride size to use when extracting patches. transformations (callable): A function/transform to apply to the data. npatches_in_row (int): The number of patches that can be extracted from a single row. npatches_in_col (int): The number of patches that can be extracted from a single column.

Examples

any_transformation()

Check if any transformations are defined.

bool

True if transformations are defined, False otherwise.

Examples

load_data(idx: int)

Load a patch from the memory-mapped file based on the given index.

idxint

Index of the patch.

dict

A dictionary containing the loaded patch data under the key ‘X’.

Examples

property shape

Returns the shape of the dataset as a tuple.

tuple

The shape of the dataset as a tuple (npatches_in_row, npatches_in_col).

Examples

transform_item(data: Dict) → Dict

Apply defined transformations to the data item.

datadict

Input data item.

dict

Transformed data item.

Examples

gunz_cm.resolution_enhancements.datasets.memmap_v2 module

Module.

Examples

class gunz_cm.resolution_enhancements.datasets.memmap_v2.MemmapDatasetV2(memmap_fpath: str, win_size: int, stride_size: int = None, transformations: Callable | None = None, max_dist: int = None)

Bases: MemmapDatasetV1

PyTorch Dataset for loading specific patches from a memory-mapped file

based on a provided max_distance attribute constraining the diagonal of the patches.

Inherits from MemmapDatasetV1 and extends its functionality as follows: - Adds a max_dist attribute to calculate the maximum diagonal length max_diag - Modifies __len__ method to return the total number of patches within the max_diag - Modifies load_data method to load a patch of size win_size with stride size stride_size

using the provided index idx only if the index corresponds to a valid pair within the max_diag

See MemmapDatasetV1 docstring for shared attributes and methods.

Examples

load_data(idx) → ndarray

Loads a patch of size win_size with stride size stride_size using the provided index idx only if the index corresponds to a valid pair within the max_diag.

idxint

The index of the patch to be loaded.

np.ndarray

A numpy array containing the patch of size win_size at the indicated valid location.

Examples

gunz_cm.resolution_enhancements.datasets.ren_memmap_v1 module

Module.

Examples

class gunz_cm.resolution_enhancements.datasets.ren_memmap_v1.RENMemmapDatasetV1(hr_memmap_fpath: str, win_size: int, stride_size: int = None, transformations: List = None, max_dist: int = None, lr_memmap_fpath: str = None, lr_ds_ratio: int = None, ret_float: bool = True)

Bases: Dataset

Dataset class for loading and manipulating memmap data.

hr_memmap_fpathstr

Path to the high-resolution memmap file.

win_sizeint

Window size for patch extraction.

stride_sizeint, optional

Stride size for patch extraction. Defaults to win_size.

transformationslist, optional

List of transformations to apply to the data. Defaults to None.

max_distint, optional

Maximum distance between patches. Defaults to shape[0].

lr_memmap_fpathstr, optional

Path to the low-resolution memmap file. Defaults to None.

lr_ds_ratioint, optional

Downscale ratio for the low-resolution data. Defaults to None.

ret_floatbool, optional

Whether to return floating-point data. Defaults to True.

_hr_fmemmap

High-resolution memmap file.

_lr_fmemmap

Low-resolution memmap file.

win_sizeint

Window size for patch extraction.

transformationslist

List of transformations to apply to the data.

stride_sizeint

Stride size for patch extraction.

npatches_in_rowint

Number of patches in each row.

npatches_in_colint

Number of patches in each column.

max_diagint

Maximum diagonal distance between patches.

i_j_pairsndarray

Array of patch indices.

ret_floatbool

Whether to return floating-point data.

Examples

any_transformation() → bool

Checks if there are any transformations present.

bool

True if there are transformations, False otherwise.

Examples

load_data(idx) → ndarray

Loads a patch of size win_size with stride size stride_size using the provided index idx.

idxint

The index of the patch to be loaded.

numpy.ndarray

A numpy array containing the patch of size win_size.

Examples

set_transformations(transformations)

Sets the transformations.

transformations

The transformations to be set.

Examples

property shape

Function shape.

Examples

Notes

transform_item(data: Dict) → Dict

Applies transformations to the data item.

datadict

Input data.

dict

Transformed data.

Examples

Module contents