tfm.vision.layers.DepthwiseConv2D

DepthwiseConv2D layer supporting CausalConv.

Inherits From: CausalConvMixin

tfm.vision.layers.DepthwiseConv2D(
    *args, use_buffered_input=False, **kwargs
)

Supports padding='causal' option (like in tf.keras.layers.Conv1D), which applies causal padding to the temporal dimension, and same padding in the spatial dimensions.

*args Arguments to be passed.
use_buffered_input A bool. If True, the input is expected to be padded beforehand. In effect, calling this layer will use 'valid' padding on the temporal dimension to simulate 'causal' padding.
**kwargs Additional keyword arguments to be passed.

use_buffered_input

Methods

call

View source

call(
    inputs
)

Calls the layer with the given inputs.

convolution_op

convolution_op(
    inputs, kernel
)