tfds.decode.SkipDecoding

Transformation which skip the decoding entirelly.

Inherits From: Decoder

tfds.decode.SkipDecoding()

Example of usage:

ds = tfds.load(
    'imagenet2012',
    split='train',
    decoders={
        'image': tfds.decode.SkipDecoding(),
    }
)

for ex in ds.take(1):
  assert ex['image'].dtype == tf.string

dtype Returns the dtype after decoding.
feature

Methods

decode_batch_example

View source

decode_batch_example(
    serialized_example
)

See FeatureConnector.decode_batch_example for details.

decode_example

View source

decode_example(
    serialized_example
)

Forward the serialized feature field.

decode_example_np

View source

decode_example_np(
    serialized_example
)

Forward the serialized feature field.

decode_ragged_example

View source

decode_ragged_example(
    serialized_example
)

See FeatureConnector.decode_ragged_example for details.

setup

View source

setup(
    *, feature
)

Transformation contructor.

The initialization of decode object is deferred because the objects only know the builder/features on which it is used after it has been constructed, the initialization is done in this function.

Args
feature tfds.features.FeatureConnector, the feature to which is applied this transformation.