unipose.datasets.BaseJointDataset

class documentation

class BaseJointDataset(Dataset, ABC):

Known subclasses: unipose.datasets.AnimalKingdomDataset, unipose.datasets.COCODataset, unipose.datasets.MPIIDataset

View In Hierarchy

Base class for datasets that have keypoints and images.

Method __getitem__ Undocumented
Method __init__ Undocumented
Method __len__ Undocumented
Method make_dataloader Make a dataloader for the dataset.
Constant EXTRA_TOKENS Extra tokens that are present in the dataset, an OrderedDict of {name: index_in_dataset}. These will be added to the batch if they are present in the mask.
Constant MAPPING Mapping from the dataset's keypoints to the unipose format, an OrderedDict of {index_in_unipose: index_in_dataset}.
Method _apply_mapping Apply the mapping from any dataset to the unipose format. This is used to reorder the keypoints and mask to the unipose format, according to the MAPPING and EXTRA_TOKENS class variables.
@abstractmethod
def __getitem__(self):
overridden in unipose.datasets.AnimalKingdomDataset, unipose.datasets.COCODataset, unipose.datasets.MPIIDataset

Undocumented

@abstractmethod
def __init__(self):
overridden in unipose.datasets.AnimalKingdomDataset, unipose.datasets.COCODataset, unipose.datasets.MPIIDataset

Undocumented

@abstractmethod
def __len__(self):
overridden in unipose.datasets.AnimalKingdomDataset, unipose.datasets.COCODataset, unipose.datasets.MPIIDataset

Undocumented

def make_dataloader(self, image_size=256, scale_factor=4, *args, **kwargs):

Make a dataloader for the dataset.

A unipose dataset returns a dictionary with the following keys: "images", "keypoint_images", "masks", "extra_keypoints", "extra_tokens".

Parameters
image_size:intThe size of the image to resize to.
scale_factor:intThe factor to scale the image by.
*argsOther arguments to pass to the DataLoader.
**kwargsOther keyword arguments to pass to the DataLoader.
EXTRA_TOKENS =
overridden in unipose.datasets.AnimalKingdomDataset, unipose.datasets.COCODataset, unipose.datasets.MPIIDataset

Extra tokens that are present in the dataset, an OrderedDict of {name: index_in_dataset}. These will be added to the batch if they are present in the mask.

Value
OrderedDict()
MAPPING =
overridden in unipose.datasets.AnimalKingdomDataset, unipose.datasets.COCODataset, unipose.datasets.MPIIDataset

Mapping from the dataset's keypoints to the unipose format, an OrderedDict of {index_in_unipose: index_in_dataset}.

Value
OrderedDict({i: i for i in range(13)})
def _apply_mapping(self, keypoints, mask):
overridden in unipose.datasets.COCODataset, unipose.datasets.MPIIDataset

Apply the mapping from any dataset to the unipose format. This is used to reorder the keypoints and mask to the unipose format, according to the MAPPING and EXTRA_TOKENS class variables.

Parameters
keypoints:torch.TensorThe keypoints to reorder.
mask:torch.TensorThe mask to reorder.