chesscog.core.dataset package¶

Module for handling datasets.

Submodules¶

chesscog.core.dataset.dataset module¶

Methods specific to handling chess datasets.

chesscog.core.dataset.dataset.build_data_loader(cfg: recap.config.CfgNode, dataset: torch.utils.data.dataset.Dataset, mode: chesscog.core.dataset.datasets.Datasets) → torch.utils.data.dataloader.DataLoader¶

Build a data loader for a dataset.

Parameters

cfg (CN) – the config object
dataset (torch.utils.data.Dataset) – the dataset
mode (Datasets) – the split

Returns

the data loader

Return type

torch.utils.data.DataLoader

chesscog.core.dataset.dataset.build_dataset(cfg: recap.config.CfgNode, mode: chesscog.core.dataset.datasets.Datasets) → torch.utils.data.dataset.Dataset¶

Build a dataset from its configuration.

Parameters

cfg (CN) – the config object
mode (Datasets) – the split (important to figure out which transforms to apply)

Returns

the dataset

Return type

torch.utils.data.Dataset

chesscog.core.dataset.dataset.color_name(color: bool) → str¶

Convert a chess color to a string.

Parameters: color (chess.Color) – the color
Returns: the string representation
Return type: str

chesscog.core.dataset.dataset.name_to_piece(name: str) → chess.Piece¶

Convert the name of a piece to an instance of chess.Piece.

Parameters: name (str) – the name of the piece
Returns: the instance of chess.Piece
Return type: chess.Piece

chesscog.core.dataset.dataset.piece_name(piece: chess.Piece) → str¶

Convert a chess piece to a string.

Parameters: piece (chess.Piece) – the piece
Returns: the corresponding string
Return type: str

chesscog.core.dataset.datasets module¶

class chesscog.core.dataset.datasets.Datasets(value)¶

Bases: enum.Enum

Enumeration of the dataset split.

TEST = 'test'¶

TRAIN = 'train'¶

VAL = 'val'¶

chesscog.core.dataset.transforms module¶

class chesscog.core.dataset.transforms.Scale(horizontal: Optional[Union[float, tuple]], vertical: Optional[Union[float, tuple]])¶

Bases: chesscog.core.dataset.transforms._HVTransform

Custom scaling transform where the horizontal and vertical scales can independently be specified.

The center of scaling is the bottom left of the image (this makes particular sense for the piece classifier).

class chesscog.core.dataset.transforms.Shear(amount: Optional[Union[tuple, float, int]])¶

Bases: object

Custom shear transform that keeps the bottom of the image invariant because for piece classification, we only want to “tilt” the top of the image.

class chesscog.core.dataset.transforms.Translate(horizontal: Optional[Union[float, tuple]], vertical: Optional[Union[float, tuple]])¶

Bases: chesscog.core.dataset.transforms._HVTransform

Custom translation transform for convenience.

chesscog.core.dataset.transforms.build_transforms(cfg: recap.config.CfgNode, mode: chesscog.core.dataset.datasets.Datasets) → Callable¶

Build the transforms for a dataset.

Parameters

cfg (CN) – the config object
mode (Datasets) – the dataset split

Returns

the transform function

Return type

typing.Callable

chesscog.core.dataset.transforms.unnormalize(x: Union[torch.Tensor, numpy.ndarray]) → Union[torch.Tensor, numpy.ndarray]¶

Unnormalize an input image.

It must be of the form ([…, W, H, 3]).

Parameters: x (typing.Union[torch.Tensor, np.ndarray]) – the input tensor/array representing the image
Returns: the unnormalized image
Return type: typing.Union[torch.Tensor, np.ndarray]