TFDS now supports the Croissant 🥐 format! Read the documentation to know more.
Module: tfds.features
Stay organized with collections
Save and categorize content based on your preferences.
API defining dataset features (image, text, scalar,...).
See the guide.
Classes
class Audio: tfds.features.FeatureConnector for audio.
class BBox: BBox(ymin, xmin, ymax, xmax)
class BBoxFeature: FeatureConnector for a normalized bounding box.
class ClassLabel: FeatureConnector for integer class labels.
class Dataset: A Dataset feature encodes a nested dataset.
class Documentation: Feature documentation such as a textual description of what this feature means.
class Encoding: Encoding type of tfds.features.Tensor.
class FeatureConnector: Abstract base class for feature types.
class FeaturesDict: Composite FeatureConnector; each feature in dict has its own connector.
class Image: FeatureConnector for images.
class LabeledImage: Image with additional label metadata.
class Scalar: A scalar feature of a particular dtype (e.g., np.int64 or np.float).
class Sequence: Composite FeatureConnector for a dict where each value is a list.
class Tensor: FeatureConnector for generic data of arbitrary shape and type.
class TensorInfo: Structure containing info on the tf.Tensor shape/dtype.
class Text: FeatureConnector for text, encoding to integers with a TextEncoder.
class Video: FeatureConnector for videos, encoding frames individually on disk.