RELAI Alpha Docs

Appearance

relai.datasets package ​

Submodules ​

relai.datasets.coco_format module ​

class relai.datasets.coco_format.CocoFormatDataset(dataset_path: str | Path, annFile: str | Path) ​

Bases: RELAIDataset

Load any object detection dataset that follows COCO data format (https://cocodataset.org/#format-data). It requires the COCO API (pycocotools) to be installed (https://github.com/cocodataset/cocoapi/tree/master).

  • Parameters:
    • dataset_path (Union *[*str , Path ]) – Root directory where images are stored.
    • annFile (Union *[*str , Path ]) – Path to the corresponding json annotation file.

relai.datasets.image_folder module ​

class relai.datasets.image_folder.ImageFolderDataset(dataset_path: str | Path) ​

Bases: RELAIClassificationDataset

Dataset wrapper for the relai_algorithms classification pipeline.

Assumes that the dataset is organized as expected by torchvision.datasets.ImageFolder as follows:

``
`

: root/dog/xxx.png root/dog/xxy.png root/dog/[…]/xxz.png
root/cat/123.png root/cat/nsdf3.png root/cat/[…]/ 

asd932_

.png 
``
`

See here for more details: https://pytorch.org/vision/stable/generated/torchvision.datasets.ImageFolder.html

  • Parameters:dataset_path (Union *[*str , Path ]) – Path to the root directory of the dataset.

property class_name_to_images : dict[str, list[int]] ​

Mapping from class name to image indices in dataset corresponding to that class.

property classes : list[str] ​

List of classes in dataset.

relai.datasets.image_folder.navigate_to_root(path: Path) ​

Navigates to the correct descendant with train/val/test folders. Assumes that there is only one descendant that is not a train/val/test folder.

relai.datasets.relai_dataset module ​

class relai.datasets.relai_dataset.RELAIClassificationDataset(*args, **kwargs) ​

Bases: RELAIDataset

property class_idx_to_images : dict[int, list[int]] ​

Mapping from class index to image indices in dataset corresponding to that class.

abstract property class_name_to_images : dict[str, list[int]] ​

Mapping from class name to image indices in dataset corresponding to that class.

property classes : list[str] ​

List of classes in dataset.

property num_classes : int ​

Number of classes in dataset.

class relai.datasets.relai_dataset.RELAIDataFormat(value, names=not given, *values, module=None, qualname=None, type=None, start=1, boundary=None) ​

Bases: StrEnum

Enum for supported dataset formats.

COCOFORMAT = 'coco_format' ​

IMAGEFOLDER = 'imagefolder' ​

class relai.datasets.relai_dataset.RELAIDataset ​

Bases: Dataset

Base class for all RELAI datasets.

class relai.datasets.relai_dataset.Subset(dataset: RELAIClassificationDataset, indices: list[int]) ​

Bases: RELAIClassificationDataset

Subset of a dataset at specified indices.

  • Parameters:
    • dataset (RELAIClassificationDataset) – The whole Dataset
    • indices (list *[*int ]) – Indices in the whole set selected for subset

property class_name_to_images : dict[str, list[int]] ​

Mapping from class name to image indices in dataset corresponding to that class.

class relai.datasets.relai_dataset.TransformDataset(dataset: RELAIDataset, transform: Callable) ​

Bases: RELAIDataset

Dataset wrapper that applies a transform to each item in the dataset.

  • Parameters:
    • dataset (RELAIDataset) – Dataset to wrap.
    • transform (Callable) – Transform to apply to each item in the dataset.

relai.datasets.segmentation_dataset module ​

class relai.datasets.segmentation_dataset.RELAISegmentationDataset(*args, **kwargs) ​

Bases: RELAIDataset

abstract get_dataset(batch_size: int) ​

Method to generate batches of dataset

Example ​

segmentation_dataset = RELAISegmentationDataset(…) for batch in segmentation_dataset.get_dataset(batch_size):

process_batch(batch)

abstract load_dataset() ​

Method to load the dataset

class relai.datasets.segmentation_dataset.RELAISegmentationDatasetCSV(dataset_path: str) ​

Bases: RELAISegmentationDataset

get_dataset(batch_size: int) ​

Method to generate batches of dataset

Example ​

segmentation_dataset = RELAISegmentationDataset(…) for batch in segmentation_dataset.get_dataset(batch_size):

process_batch(batch)

load_dataset() ​

Method to load the dataset

class relai.datasets.segmentation_dataset.SegmentationDataPair(input_image_path, input_image, ground_truth_path, ground_truth) ​

Bases: tuple

ground_truth ​

Alias for field number 3

ground_truth_path ​

Alias for field number 2

input_image ​

Alias for field number 1

input_image_path ​

Alias for field number 0

relai.datasets.utils module ​

relai.datasets.utils.coco_collate(batch) ​

relai.datasets.utils.get_dataset(data_format: RELAIDataFormat, dataset_path: str | Path, *dataset_args, **dataset_kwargs) ​

Returns a RELAIDataset object based on the specified format and path.

  • Parameters:
    • dataset_format (RELAIDataFormat) – Format of dataset.
    • dataset_path (Union *[*str , Path ]) – Path to dataset.
    • **dataset_args – Positional arguments to pass to dataset constructor.
    • **dataset_kwargs – Keyword arguments to pass to dataset constructor.
  • Returns: Dataset object.
  • Return type:RELAIDataset

Module contents ​