|TensorFlow 1 version||View source on GitHub|
The transpose of
tf.nn.atrous_conv2d_transpose( value, filters, output_shape, rate, padding, name=None )
This operation is sometimes called "deconvolution" after Deconvolutional
but is really the transpose (gradient) of
atrous_conv2d rather than an
value: A 4-D
float. It needs to be in the default
NHWCformat. Its shape is
[batch, in_height, in_width, in_channels].
filters: A 4-D
Tensorwith the same type as
[filter_height, filter_width, out_channels, in_channels].
in_channelsdimension must match that of
value. Atrous convolution is equivalent to standard convolution with upsampled filters with effective height
filter_height + (filter_height - 1) * (rate - 1)and effective width
filter_width + (filter_width - 1) * (rate - 1), produced by inserting
rate - 1zeros along consecutive elements across the
filters' spatial dimensions.
output_shape: A 1-D
Tensorof shape representing the output shape of the deconvolution op.
rate: A positive int32. The stride with which we sample input values across the
widthdimensions. Equivalently, the rate by which we upsample the filter values by inserting zeros across the
widthdimensions. In the literature, the same parameter is sometimes called
padding: A string, either
'SAME'. The padding algorithm.
name: Optional name for the returned tensor.
Tensor with the same type as
ValueError: If input/output depth does not match
filters' shape, or if padding is other than
'SAME', or if the
rateis less than one, or if the output_shape is not a tensor with 4 elements.