/// <summary>
/// Computes a 2-D convolution given 4-D `input` and `filter` tensors.
///
/// Given an input tensor of shape `[batch, in_height, in_width, in_channels]`
/// and a filter / kernel tensor of shape
/// `[filter_height, filter_width, in_channels, out_channels]`, this op
/// performs the following:
///
/// 1. Flattens the filter to a 2-D matrix with shape
/// `[filter_height * filter_width * in_channels, output_channels]`.
/// 2. Extracts image patches from the input tensor to form a *virtual*
/// tensor of shape `[batch, out_height, out_width,
/// filter_height * filter_width * in_channels]`.
/// 3. For each patch, right-multiplies the filter matrix and the image patch
/// vector.
/// </summary>
/// <param name="parameters"></param>
/// <returns></returns>
public static Tensor conv2d(Conv2dParams parameters)
{
if (tf.executing_eagerly())
{
var results = tf.Runner.TFE_FastPathExecute(tf.Context, tf.Context.DeviceName,
"Conv2D", parameters.Name,
null,
parameters.Input, parameters.Filter,
"strides", parameters.Strides,
"use_cudnn_on_gpu", parameters.UseCudnnOnGpu,
"padding", parameters.Padding,
"explicit_paddings", parameters.ExplicitPaddings,
"data_format", parameters.DataFormat,
"dilations", parameters.Dilations);
return(results[0]);
}
var _op = tf.OpDefLib._apply_op_helper("Conv2D", name: parameters.Name, args: new
{
input = parameters.Input,
filter = parameters.Filter,
strides = parameters.Strides,
padding = parameters.Padding,
use_cudnn_on_gpu = parameters.UseCudnnOnGpu,
explicit_paddings = parameters.ExplicitPaddings,
data_format = parameters.DataFormat,
dilations = parameters.Dilations
});
return(_op.outputs[0]);
}
/// <summary>
/// Computes a 2-D convolution given 4-D `input` and `filter` tensors.
///
/// Given an input tensor of shape `[batch, in_height, in_width, in_channels]`
/// and a filter / kernel tensor of shape
/// `[filter_height, filter_width, in_channels, out_channels]`, this op
/// performs the following:
///
/// 1. Flattens the filter to a 2-D matrix with shape
/// `[filter_height * filter_width * in_channels, output_channels]`.
/// 2. Extracts image patches from the input tensor to form a *virtual*
/// tensor of shape `[batch, out_height, out_width,
/// filter_height * filter_width * in_channels]`.
/// 3. For each patch, right-multiplies the filter matrix and the image patch
/// vector.
/// </summary>
/// <param name="parameters"></param>
/// <returns></returns>
public static Tensor conv2d(Conv2dParams parameters)
=> tf.Context.ExecuteOp("Conv2D", parameters.Name, new ExecuteOpArgs(parameters.Input, parameters.Filter)
.SetAttributes(new
{
strides = parameters.Strides,
padding = parameters.Padding,
use_cudnn_on_gpu = parameters.UseCudnnOnGpu,
explicit_paddings = parameters.ExplicitPaddings,
data_format = parameters.DataFormat,
dilations = parameters.Dilations
}));
/// <summary>
/// Computes a 2-D convolution given 4-D `input` and `filter` tensors.
///
/// Given an input tensor of shape `[batch, in_height, in_width, in_channels]`
/// and a filter / kernel tensor of shape
/// `[filter_height, filter_width, in_channels, out_channels]`, this op
/// performs the following:
///
/// 1. Flattens the filter to a 2-D matrix with shape
/// `[filter_height * filter_width * in_channels, output_channels]`.
/// 2. Extracts image patches from the input tensor to form a *virtual*
/// tensor of shape `[batch, out_height, out_width,
/// filter_height * filter_width * in_channels]`.
/// 3. For each patch, right-multiplies the filter matrix and the image patch
/// vector.
/// </summary>
/// <param name="parameters"></param>
/// <returns></returns>
public static Tensor conv2d(Conv2dParams parameters)
{
var _op = _op_def_lib._apply_op_helper("Conv2D", name: parameters.Name, args: new
{
input = parameters.Input,
filter = parameters.Filter,
strides = parameters.Strides,
padding = parameters.Padding,
use_cudnn_on_gpu = parameters.UseCudnnOnGpu,
explicit_paddings = parameters.ExplicitPaddings,
data_format = parameters.DataFormat,
dilations = parameters.Dilations
});
return(_op.outputs[0]);
}
/// <summary>
/// Computes the gradients of convolution with respect to the input.
/// </summary>
/// <param name="parameters"></param>
/// <returns></returns>
public static Tensor conv2d_backprop_input(Conv2dParams parameters)
{
var _op = tf.OpDefLib._apply_op_helper("Conv2DBackpropInput", name: parameters.Name, args: new
{
input_sizes = parameters.InputSizes,
filter = parameters.Filter,
out_backprop = parameters.OutBackProp,
strides = parameters.Strides,
padding = parameters.Padding,
use_cudnn_on_gpu = parameters.UseCudnnOnGpu,
explicit_paddings = parameters.ExplicitPaddings,
data_format = parameters.DataFormat,
dilations = parameters.Dilations
});
return(_op.outputs[0]);
}
