fruit_project.utils.data
========================

.. py:module:: fruit_project.utils.data


Functions
---------

.. autoapisummary::

   fruit_project.utils.data.make_datasets
   fruit_project.utils.data.get_sampler
   fruit_project.utils.data.make_dataloaders
   fruit_project.utils.data.get_labels_and_mappings
   fruit_project.utils.data.collate_fn
   fruit_project.utils.data.set_transforms


Module Contents
---------------

.. py:function:: make_datasets(cfg: omegaconf.DictConfig) -> Tuple[fruit_project.utils.datasets.det_dataset.DET_DS, fruit_project.utils.datasets.det_dataset.DET_DS, fruit_project.utils.datasets.det_dataset.DET_DS]

   Creates training, testing, and validation datasets.

   :param cfg: Configuration object containing dataset parameters.
   :type cfg: DictConfig

   :returns: The training, testing, and validation datasets.
   :rtype: Tuple[DET_DS, DET_DS, DET_DS]


.. py:function:: get_sampler(train_ds: fruit_project.utils.datasets.det_dataset.DET_DS, generator) -> torch.utils.data.WeightedRandomSampler

   Creates a WeightedRandomSampler for the training dataset.
   Handles the new dataset format which returns a single dictionary.


.. py:function:: make_dataloaders(cfg: omegaconf.DictConfig, train_ds_base: fruit_project.utils.datasets.det_dataset.DET_DS, test_ds: fruit_project.utils.datasets.det_dataset.DET_DS, val_ds: fruit_project.utils.datasets.det_dataset.DET_DS, generator: torch.Generator, processor: transformers.AutoImageProcessor, transforms: albumentations.Compose) -> Tuple[torch.utils.data.DataLoader, torch.utils.data.DataLoader, torch.utils.data.DataLoader, Dict]

   Creates dataloaders for training, testing, and validation datasets.

   :param cfg: Configuration object containing dataloader parameters.
   :type cfg: DictConfig
   :param train_ds: The training dataset.
   :type train_ds: DET_DS
   :param test_ds: The testing dataset.
   :type test_ds: DET_DS
   :param val_ds: The validation dataset.
   :type val_ds: DET_DS
   :param generator: A PyTorch generator for reproducibility.
   :type generator: torch.Generator
   :param processor: Processor for image preprocessing.
   :type processor: AutoImageProcessor
   :param transforms: Transformations to apply to the datasets.
   :type transforms: Compose

   :returns: The training, testing, validation dataloaders and a training sample.
   :rtype: Tuple[DataLoader, DataLoader, DataLoader, Tuple[torch.Tensor, torch.Tensor]]


.. py:function:: get_labels_and_mappings(train_labels: List, test_labels: List) -> Tuple[List, Dict, Dict]

   Generates labels and mappings for class IDs and names.

   :param train_labels: List of labels from the training dataset.
   :type train_labels: List
   :param test_labels: List of labels from the testing dataset.
   :type test_labels: List

   :returns:

             A tuple containing:
                 - labels (List): Sorted list of unique labels.
                 - id2lbl (Dict): Mapping from class IDs to labels.
                 - lbl2id (Dict): Mapping from labels to class IDs.
   :rtype: Tuple[List, Dict, Dict]


.. py:function:: collate_fn(batch: transformers.BatchEncoding, model_name) -> Dict

   Collates a batch of data for the dataloader.

   :param batch: A batch of data containing images and targets.
   :type batch: BatchEncoding

   :returns: Processed batch and list of targets.
   :rtype: Tuple[BatchEncoding, List]


.. py:function:: set_transforms(train_ds, test_ds, val_ds, transforms: albumentations.Compose)

   Sets transformations for the datasets in the dataloaders.

   :param train_dl: Training dataloader.
   :type train_dl: DataLoader
   :param test_dl: Testing dataloader.
   :type test_dl: DataLoader
   :param val_dl: Validation dataloader.
   :type val_dl: DataLoader
   :param transforms: Transformations to apply.
   :type transforms: Compose

   :returns: Updated dataloaders with transformations applied.
   :rtype: Tuple[DataLoader, DataLoader, DataLoader]