|
|
|
|
@ -11067,8 +11067,9 @@ def soft_relu(x, threshold=40.0, name=None):
|
|
|
|
|
|
|
|
|
|
def flatten(x, axis=1, name=None):
|
|
|
|
|
"""
|
|
|
|
|
**Flatten layer**
|
|
|
|
|
Flattens the input tensor into a 2D matrix.
|
|
|
|
|
**Flatten op**
|
|
|
|
|
|
|
|
|
|
Flatten the input tensor into a 2D matrix.
|
|
|
|
|
|
|
|
|
|
For Example:
|
|
|
|
|
|
|
|
|
|
@ -11097,21 +11098,20 @@ def flatten(x, axis=1, name=None):
|
|
|
|
|
Out.shape = (1, 3 * 100 * 100 * 4)
|
|
|
|
|
|
|
|
|
|
Args:
|
|
|
|
|
x (Variable): A tensor of rank >= axis.
|
|
|
|
|
x (Variable): A tensor of rank >= axis. A tensor with type float32,
|
|
|
|
|
float64, int8, int32, int64.
|
|
|
|
|
axis (int): Indicate up to which input dimensions (exclusive) should
|
|
|
|
|
be flattened to the outer dimension of the output.
|
|
|
|
|
The value for axis must be in the range [0, R], where R
|
|
|
|
|
is the rank of the input tensor. When axis = 0, the shape
|
|
|
|
|
of the output tensor is (1, (d_0 X d_1 ... d_n), where the
|
|
|
|
|
shape of the input tensor is (d_0, d_1, ... d_n).
|
|
|
|
|
name(str|None): A name for this layer(optional). If set None, the layer
|
|
|
|
|
will be named automatically.
|
|
|
|
|
is the rank of the input tensor. Default: 1.
|
|
|
|
|
name(str, Optional): For details, please refer to :ref:`api_guide_Name`.
|
|
|
|
|
Generally, no setting is required. Default: None.
|
|
|
|
|
|
|
|
|
|
Returns:
|
|
|
|
|
Variable: A 2D tensor with the contents of the input tensor, with input \
|
|
|
|
|
dimensions up to axis flattened to the outer dimension of \
|
|
|
|
|
the output and remaining input dimensions flattened into the \
|
|
|
|
|
inner dimension of the output.
|
|
|
|
|
inner dimension of the output. A Tensor with type same as input x.
|
|
|
|
|
|
|
|
|
|
Raises:
|
|
|
|
|
ValueError: If x is not a variable.
|
|
|
|
|
@ -11122,8 +11122,10 @@ def flatten(x, axis=1, name=None):
|
|
|
|
|
.. code-block:: python
|
|
|
|
|
|
|
|
|
|
import paddle.fluid as fluid
|
|
|
|
|
x = fluid.layers.data(name="x", shape=[4, 4, 3], dtype="float32")
|
|
|
|
|
x = fluid.layers.data(name="x", shape=[4, 4, 3], append_batch_size=False, dtype="float32")
|
|
|
|
|
# x shape is [4, 4, 3]
|
|
|
|
|
out = fluid.layers.flatten(x=x, axis=2)
|
|
|
|
|
# out shape is [16, 3]
|
|
|
|
|
"""
|
|
|
|
|
helper = LayerHelper('flatten', **locals())
|
|
|
|
|
|
|
|
|
|
@ -14618,7 +14620,7 @@ def fsp_matrix(x, y):
|
|
|
|
|
|
|
|
|
|
**FSP matrix op**
|
|
|
|
|
|
|
|
|
|
This op is used to calculate the flow of solution procedure (FSP) matrix of two feature maps.
|
|
|
|
|
This op is used to calculate the flow of solution procedure (FSP) matrix of two 4-D Tensor feature maps.
|
|
|
|
|
Given feature map x with shape [x_channel, h, w] and feature map y with shape
|
|
|
|
|
[y_channel, h, w], we can get the fsp matrix of x and y in two steps:
|
|
|
|
|
|
|
|
|
|
@ -14630,15 +14632,18 @@ def fsp_matrix(x, y):
|
|
|
|
|
|
|
|
|
|
Args:
|
|
|
|
|
|
|
|
|
|
x (Variable): A feature map with shape [batch_size, x_channel, height, width].
|
|
|
|
|
y (Variable): A feature map with shape [batch_size, y_channel, height, width].
|
|
|
|
|
x (Variable): A 4-D Tensor feature map with shape [batch_size, x_channel, height, width].
|
|
|
|
|
A Tensor with type float32, float64.
|
|
|
|
|
y (Variable): A 4-D Tensor feature map with shape [batch_size, y_channel, height, width].
|
|
|
|
|
The y_channel can be different with the x_channel of Input(X)
|
|
|
|
|
while the other dimensions must be the same with Input(X)'s.
|
|
|
|
|
while the other dimensions must be the same with Input(X)'s. A Tensor with
|
|
|
|
|
type float32, float64.
|
|
|
|
|
|
|
|
|
|
Returns:
|
|
|
|
|
|
|
|
|
|
fsp matrix (Variable): The output of FSP op with shape [batch_size, x_channel, y_channel].
|
|
|
|
|
The x_channel is the channel of x and the y_channel is the channel of y.
|
|
|
|
|
The x_channel is the channel of x and the y_channel is the channel of y. A Tensor with
|
|
|
|
|
type float32, float64.
|
|
|
|
|
|
|
|
|
|
Examples:
|
|
|
|
|
|
|
|
|
|
@ -14900,7 +14905,7 @@ def deformable_conv(input,
|
|
|
|
|
modulated=True,
|
|
|
|
|
name=None):
|
|
|
|
|
"""
|
|
|
|
|
**Deformable Convolution Layer**
|
|
|
|
|
**Deformable Convolution op**
|
|
|
|
|
|
|
|
|
|
Compute 2-D deformable convolution on 4-D input.
|
|
|
|
|
Given input image x, output feature map y, the deformable convolution operation can be expressed as follow:
|
|
|
|
|
@ -14919,7 +14924,7 @@ def deformable_conv(input,
|
|
|
|
|
y(p) = \sum_{k=1}^{K}{w_k * x(p + p_k + \Delta p_k)}
|
|
|
|
|
|
|
|
|
|
Where :math:`\Delta p_k` and :math:`\Delta m_k` are the learnable offset and modulation scalar for the k-th location,
|
|
|
|
|
which :math:`\Delta m_k` is one in deformable convolution v1. Please refer to `Deformable ConvNets v2: More Deformable, Better Results
|
|
|
|
|
Which :math:`\Delta m_k` is one in deformable convolution v1. Please refer to `Deformable ConvNets v2: More Deformable, Better Results
|
|
|
|
|
<https://arxiv.org/abs/1811.11168v2>`_ and `Deformable Convolutional Networks <https://arxiv.org/abs/1703.06211>`_.
|
|
|
|
|
|
|
|
|
|
Example:
|
|
|
|
|
@ -14945,12 +14950,16 @@ def deformable_conv(input,
|
|
|
|
|
W_{out}&= \\frac{(W_{in} + 2 * paddings[1] - (dilations[1] * (W_f - 1) + 1))}{strides[1]} + 1
|
|
|
|
|
|
|
|
|
|
Args:
|
|
|
|
|
input (Variable): The input image with [N, C, H, W] format.
|
|
|
|
|
input (Variable): The input image with [N, C, H, W] format. A Tensor with type
|
|
|
|
|
float32, float64.
|
|
|
|
|
offset (Variable): The input coordinate offset of deformable convolution layer.
|
|
|
|
|
Mask (Variable): The input mask of deformable covolution layer.
|
|
|
|
|
A Tensor with type float32, float64.
|
|
|
|
|
Mask (Variable, Optional): The input mask of deformable covolution layer.
|
|
|
|
|
A Tensor with type float32, float64.It should be None when you use
|
|
|
|
|
deformable_conv_v2.
|
|
|
|
|
num_filters(int): The number of filter. It is as same as the output
|
|
|
|
|
image channel.
|
|
|
|
|
filter_size (int|tuple|None): The filter size. If filter_size is a tuple,
|
|
|
|
|
filter_size (int|tuple): The filter size. If filter_size is a tuple,
|
|
|
|
|
it must contain two integers, (filter_size_H, filter_size_W).
|
|
|
|
|
Otherwise, the filter will be a square.
|
|
|
|
|
stride (int|tuple): The stride size. If stride is a tuple, it must
|
|
|
|
|
@ -14974,24 +14983,24 @@ def deformable_conv(input,
|
|
|
|
|
than this value; if you face out of memory problem, you can try
|
|
|
|
|
to use a smaller value here.
|
|
|
|
|
Default: im2col_step = 64.
|
|
|
|
|
param_attr (ParamAttr|None): The parameter attribute for learnable parameters/weights
|
|
|
|
|
param_attr (ParamAttr, Optional): The parameter attribute for learnable parameters/weights
|
|
|
|
|
of deformable conv. If it is set to None or one attribute of ParamAttr,
|
|
|
|
|
deformable conv will create ParamAttr as param_attr.
|
|
|
|
|
If the Initializer of the param_attr is not set, the parameter is
|
|
|
|
|
initialized with :math:`Normal(0.0, std)`, and the
|
|
|
|
|
:math:`std` is :math:`(\\frac{2.0 }{filter\_elem\_num})^{0.5}`. Default: None.
|
|
|
|
|
bias_attr (ParamAttr|bool|None): The parameter attribute for the bias of
|
|
|
|
|
bias_attr (ParamAttr|bool, Optional): The parameter attribute for the bias of
|
|
|
|
|
deformable conv layer. If it is set to False, no bias will be added
|
|
|
|
|
to the output units. If it is set to None or one attribute of ParamAttr, conv2d
|
|
|
|
|
will create ParamAttr as bias_attr. If the Initializer of the bias_attr
|
|
|
|
|
is not set, the bias is initialized zero. Default: None.
|
|
|
|
|
modulated (bool): Make sure which version should be used between v1 and v2, where v2 is \
|
|
|
|
|
used while True. Default: True.
|
|
|
|
|
name (str|None): A name for this layer(optional). If set None, the layer
|
|
|
|
|
will be named automatically. Default: None
|
|
|
|
|
name(str, Optional): For details, please refer to :ref:`api_guide_Name`.
|
|
|
|
|
Generally, no setting is required. Default: None.
|
|
|
|
|
Returns:
|
|
|
|
|
Variable: The tensor variable storing the deformable convolution \
|
|
|
|
|
result.
|
|
|
|
|
result. A Tensor with type float32, float64.
|
|
|
|
|
Raises:
|
|
|
|
|
ValueError: If the shapes of input, filter_size, stride, padding and
|
|
|
|
|
groups mismatch.
|
|
|
|
|
@ -15001,19 +15010,23 @@ def deformable_conv(input,
|
|
|
|
|
#deformable conv v2:
|
|
|
|
|
|
|
|
|
|
import paddle.fluid as fluid
|
|
|
|
|
data = fluid.layers.data(name='data', shape=[3, 32, 32], dtype='float32')
|
|
|
|
|
offset = fluid.layers.data(name='offset', shape=[18, 32, 32], dtype='float32')
|
|
|
|
|
mask = fluid.layers.data(name='mask', shape=[9, 32, 32], dtype='float32')
|
|
|
|
|
C_in, H_in, W_in = 3, 32, 32
|
|
|
|
|
filter_size, deformable_groups = 3, 1
|
|
|
|
|
data = fluid.layers.data(name='data', shape=[C_in, H_in, W_in], dtype='float32')
|
|
|
|
|
offset = fluid.layers.data(name='offset', shape=[2*deformable_groups*filter_size**2, H_in, W_in], dtype='float32')
|
|
|
|
|
mask = fluid.layers.data(name='mask', shape=[deformable_groups*filter_size**2, H_in, W_in], dtype='float32')
|
|
|
|
|
out = fluid.layers.deformable_conv(input=data, offset=offset, mask=mask,
|
|
|
|
|
num_filters=2, filter_size=3, padding=1, modulated=True)
|
|
|
|
|
num_filters=2, filter_size=filter_size, padding=1, modulated=True)
|
|
|
|
|
|
|
|
|
|
#deformable conv v1:
|
|
|
|
|
|
|
|
|
|
import paddle.fluid as fluid
|
|
|
|
|
data = fluid.layers.data(name='data', shape=[3, 32, 32], dtype='float32')
|
|
|
|
|
offset = fluid.layers.data(name='offset', shape=[18, 32, 32], dtype='float32')
|
|
|
|
|
C_in, H_in, W_in = 3, 32, 32
|
|
|
|
|
filter_size, deformable_groups = 3, 1
|
|
|
|
|
data = fluid.layers.data(name='data', shape=[C_in, H_in, W_in], dtype='float32')
|
|
|
|
|
offset = fluid.layers.data(name='offset', shape=[2*deformable_groups*filter_size**2, H_in, W_in], dtype='float32')
|
|
|
|
|
out = fluid.layers.deformable_conv(input=data, offset=offset, mask=None,
|
|
|
|
|
num_filters=2, filter_size=3, padding=1, modulated=False)
|
|
|
|
|
num_filters=2, filter_size=filter_size, padding=1, modulated=False)
|
|
|
|
|
"""
|
|
|
|
|
|
|
|
|
|
num_channels = input.shape[1]
|
|
|
|
|
|
Bai Yifan
6 years ago
committed by
GitHub