几个paddle组件
Conv2D
二维卷积层
https://www.paddlepaddle.org.cn/documentation/docs/zh/api/paddle/nn/Conv2D_cn.html
- in_channels (int) - 输入图像的通道数。
- out_channels (int) - 由卷积操作产生的输出的通道数。
- kernel_size (int|list|tuple) - 卷积核大小。
- stride (int|list|tuple，可选) - 步长大小。默认值：1。
- padding (int|list|tuple|str，可选) - 填充大小。默认值：0。
- dilation (int|list|tuple，可选) - 空洞大小。默认值：1。
[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-MVW1RfH2-1656773613140)(F:\typora\images\image-20220702202001345.png)]
- bias_attr （ParamAttr|bool，可选）- 指定偏置参数属性的对象。若
bias_attr为bool类型，只支持为False，表示没有偏置参数。默认值为None，表示使用默认的偏置参数属性。
BatchNorm2D
批归一化层
num_features (int) - 指明输入 Tensor 的通道数量。
下采样
对于一幅图像I尺寸为MN，对其进行s倍下采样，即得到(M/s)(N/s)尺寸的得分辨率图像，当然s应该是M和N的公约数才行，如果考虑的是矩阵形式的图像，就是把原始图像s*s窗口内的图像变成一个像素，这个像素点的值就是窗口内所有像素的均值：
BasicBlock与BottleneckBlock
BasicBlock和Bottleneck的区别在于前者是用两个3x3的卷积组成的，后者是用两个1x1的卷积加一个3x3的卷积组成的。
# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import division
from __future__ import print_function
import paddle
import paddle.nn as nn
from paddle.utils.download import get_weights_path_from_url
__all__ = []
model_urls = {
'resnet18': ('https://paddle-hapi.bj.bcebos.com/models/resnet18.pdparams',
'cf548f46534aa3560945be4b95cd11c4'),
'resnet34': ('https://paddle-hapi.bj.bcebos.com/models/resnet34.pdparams',
'8d2275cf8706028345f78ac0e1d31969'),
'resnet50': ('https://paddle-hapi.bj.bcebos.com/models/resnet50.pdparams',
'ca6f485ee1ab0492d38f323885b0ad80'),
'resnet101': ('https://paddle-hapi.bj.bcebos.com/models/resnet101.pdparams',
'02f35f034ca3858e1e54d4036443c92d'),
'resnet152': ('https://paddle-hapi.bj.bcebos.com/models/resnet152.pdparams',
'7ad16a2f1e7333859ff986138630fd7a'),
'resnext50_32x4d':
('https://paddle-hapi.bj.bcebos.com/models/resnext50_32x4d.pdparams',
'dc47483169be7d6f018fcbb7baf8775d'),
"resnext50_64x4d":
('https://paddle-hapi.bj.bcebos.com/models/resnext50_64x4d.pdparams',
'063d4b483e12b06388529450ad7576db'),
'resnext101_32x4d':
('https://paddle-hapi.bj.bcebos.com/models/resnext101_32x4d.pdparams',
'967b090039f9de2c8d06fe994fb9095f'),
'resnext101_64x4d':
('https://paddle-hapi.bj.bcebos.com/models/resnext101_64x4d.pdparams',
'98e04e7ca616a066699230d769d03008'),
'resnext152_32x4d':
('https://paddle-hapi.bj.bcebos.com/models/resnext152_32x4d.pdparams',
'18ff0beee21f2efc99c4b31786107121'),
'resnext152_64x4d':
('https://paddle-hapi.bj.bcebos.com/models/resnext152_64x4d.pdparams',
'77c4af00ca42c405fa7f841841959379'),
'wide_resnet50_2':
('https://paddle-hapi.bj.bcebos.com/models/wide_resnet50_2.pdparams',
'0282f804d73debdab289bd9fea3fa6dc'),
'wide_resnet101_2':
('https://paddle-hapi.bj.bcebos.com/models/wide_resnet101_2.pdparams',
'd4360a2d23657f059216f5d5a1a9ac93'),
}
class BasicBlock(nn.Layer):
expansion = 1#不改变深度
def __init__(self,
inplanes,
planes,#“基准通道数”，不是输出通道数
stride=1,
downsample=None,#下采样, 每隔n个间隔进行抽取，output.shape长宽变成1/2
groups=1,
base_width=64,
dilation=1,
norm_layer=None):
super(BasicBlock, self).__init__()
if norm_layer is None:
norm_layer = nn.BatchNorm2D#批归一化
if dilation > 1:
raise NotImplementedError(
"Dilation > 1 not supported in BasicBlock")
self.conv1 = nn.Conv2D(inplanes,#输入通道数
planes,#输出通道数
3,#卷积核大小。
padding=1,
stride=stride,
bias_attr=False)#表示没有偏执参数
self.bn1 = norm_layer(planes)#批归一化
self.relu = nn.ReLU()#激活
self.conv2 = nn.Conv2D(planes, planes, 3, padding=1, bias_attr=False)
self.bn2 = norm_layer(planes)#批归一化
self.downsample = downsample#是否进行下采样的判断
self.stride = stride
def forward(self, x):
#将输出特征直接设置为的输入
identity = x
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
#常规通过卷积获取输出特征
if self.downsample is not None:
identity = self.downsample(x)
#最后的特征值=输入X+正常卷积获取的输出out
out += identity
#激活
out = self.relu(out)
return out
#输入channel(通道)从大变小, 再从小变大。
class BottleneckBlock(nn.Layer):
expansion = 4#最后一次卷积将深度扩大4倍
def __init__(self,
inplanes,
planes,
stride=1,
downsample=None,
groups=1,#将基准的通道数扩张groups倍
base_width=64,#基准的通道数默认为64
dilation=1,
norm_layer=None):
super(BottleneckBlock, self).__init__()
if norm_layer is None:
norm_layer = nn.BatchNorm2D
width = int(planes * (base_width / 64.)) * groups#将基准的通道数扩张groups倍
self.conv1 = nn.Conv2D(inplanes, width, 1, bias_attr=False)#输入通道数inplanes, 输出通道数width,卷积核大小为1*1,
self.bn1 = norm_layer(width)
self.conv2 = nn.Conv2D(width,#输入通道数width
width,#输出通道数width
3,#卷积核大小3*3
padding=dilation,
stride=stride,
groups=groups,
dilation=dilation,
bias_attr=False)
self.bn2 = norm_layer(width)
self.conv3 = nn.Conv2D(width,
planes * self.expansion,#最后一次卷积将深度扩大4倍
1,#卷积核大小为1*1,
bias_attr=False)
self.bn3 = norm_layer(planes * self.expansion)
self.relu = nn.ReLU()
self.downsample = downsample
self.stride = stride
def forward(self, x):
identity = x
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
out = self.relu(out)
out = self.conv3(out)
out = self.bn3(out)
if self.downsample is not None:
identity = self.downsample(x)
out += identity
out = self.relu(out)
return out
class ResNet(nn.Layer):
"""ResNet model from
`"Deep Residual Learning for Image Recognition" <https://arxiv.org/pdf/1512.03385.pdf>`_
Args:
Block (BasicBlock|BottleneckBlock): block module of model.
depth (int, optional): layers of ResNet, Default: 50.
width (int, optional): base width per convolution group for each convolution block, Default: 64.
num_classes (int, optional): output dim of last fc layer. If num_classes <=0, last fc layer
will not be defined. Default: 1000.
with_pool (bool, optional): use pool before the last fc layer or not. Default: True.
groups (int, optional): number of groups for each convolution block, Default: 1.
Returns:
ResNet model. An instance of :ref:`api_fluid_dygraph_Layer`.
Examples:
.. code-block:: python
import paddle
from paddle.vision.models import ResNet
from paddle.vision.models.resnet import BottleneckBlock, BasicBlock
# build ResNet with 18 layers
resnet18 = ResNet(BasicBlock, 18)
# build ResNet with 50 layers
resnet50 = ResNet(BottleneckBlock, 50)
# build Wide ResNet model
wide_resnet50_2 = ResNet(BottleneckBlock, 50, width=64*2)
# build ResNeXt model
resnext50_32x4d = ResNet(BottleneckBlock, 50, width=4, groups=32)
x = paddle.rand([1, 3, 224, 224])
out = resnet18(x)
print(out.shape)
# [1, 1000]
"""
def __init__(self,
block,#使用的block,bottleneck或者basicblock
depth=50,#层数
width=64,#卷积核数量
num_classes=1000,#fc层的参数数量
with_pool=True,
groups=1):
super(ResNet, self).__init__()
layer_cfg = {
18: [2, 2, 2, 2],#8*2+2 basicblock每个两层
34: [3, 4, 6, 3],#16*2+2
50: [3, 4, 6, 3],#16*3+2 bottleblock每个3层
101: [3, 4, 23, 3],#33*3+2
152: [3, 8, 36, 3]#50*3+2
}
layers = layer_cfg[depth]
self.groups = groups
self.base_width = width
self.num_classes = num_classes
self.with_pool = with_pool
self._norm_layer = nn.BatchNorm2D
self.inplanes = 64
self.dilation = 1
self.conv1 = nn.Conv2D(3,
self.inplanes,
kernel_size=7,
stride=2,
padding=3,
bias_attr=False)
self.bn1 = self._norm_layer(self.inplanes)
self.relu = nn.ReLU()
self.maxpool = nn.MaxPool2D(kernel_size=3, stride=2, padding=1)
self.layer1 = self._make_layer(block, 64, layers[0])
self.layer2 = self._make_layer(block, 128, layers[1], stride=2)
self.layer3 = self._make_layer(block, 256, layers[2], stride=2)
self.layer4 = self._make_layer(block, 512, layers[3], stride=2)
if with_pool:
self.avgpool = nn.AdaptiveAvgPool2D((1, 1))
if num_classes > 0:
self.fc = nn.Linear(512 * block.expansion, num_classes)
#根据输入参数创建对应的层结构
def _make_layer(self, block, planes, blocks, stride=1, dilate=False):
norm_layer = self._norm_layer
downsample = None
previous_dilation = self.dilation
if dilate:
self.dilation *= stride
stride = 1
if stride != 1 or self.inplanes != planes * block.expansion:
downsample = nn.Sequential(
nn.Conv2D(self.inplanes,
planes * block.expansion,
1,
stride=stride,
bias_attr=False),
norm_layer(planes * block.expansion),
)
layers = []
layers.append(
block(self.inplanes, planes, stride, downsample, self.groups,
self.base_width, previous_dilation, norm_layer))
self.inplanes = planes * block.expansion
for _ in range(1, blocks):
layers.append(
block(self.inplanes,
planes,
groups=self.groups,
base_width=self.base_width,
norm_layer=norm_layer))
return nn.Sequential(*layers)
def forward(self, x):
x = self.conv1(x)#这里算一层
x = self.bn1(x)
x = self.relu(x)
x = self.maxpool(x)
x = self.layer1(x)
x = self.layer2(x)
x = self.layer3(x)
x = self.layer4(x)
#这里算一层
if self.with_pool:
x = self.avgpool(x)
if self.num_classes > 0:
x = paddle.flatten(x, 1)
x = self.fc(x)
return x
def _resnet(arch, Block, depth, pretrained, **kwargs):
model = ResNet(Block, depth, **kwargs)
if pretrained:
assert arch in model_urls, "{} model do not have a pretrained model now, you should set pretrained=False".format(
arch)
weight_path = get_weights_path_from_url(model_urls[arch][0],
model_urls[arch][1])
param = paddle.load(weight_path)
model.set_dict(param)
return model
def resnet18(pretrained=False, **kwargs):
"""ResNet 18-layer model from
`"Deep Residual Learning for Image Recognition" <https://arxiv.org/pdf/1512.03385.pdf>`_
Args:
pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained
on ImageNet. Default: False.
Returns:
ResNet 18-layer model. An instance of :ref:`api_fluid_dygraph_Layer`.
Examples:
.. code-block:: python
import paddle
from paddle.vision.models import resnet18
# build model
model = resnet18()
# build model and load imagenet pretrained weight
# model = resnet18(pretrained=True)
x = paddle.rand([1, 3, 224, 224])
out = model(x)
print(out.shape)
# [1, 1000]
"""
return _resnet('resnet18', BasicBlock, 18, pretrained, **kwargs)
def resnet34(pretrained=False, **kwargs):
"""ResNet 34-layer model from
`"Deep Residual Learning for Image Recognition" <https://arxiv.org/pdf/1512.03385.pdf>`_
Args:
pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained
on ImageNet. Default: False.
Returns:
ResNet 34-layer model. An instance of :ref:`api_fluid_dygraph_Layer`.
Examples:
.. code-block:: python
import paddle
from paddle.vision.models import resnet34
# build model
model = resnet34()
# build model and load imagenet pretrained weight
# model = resnet34(pretrained=True)
x = paddle.rand([1, 3, 224, 224])
out = model(x)
print(out.shape)
# [1, 1000]
"""
return _resnet('resnet34', BasicBlock, 34, pretrained, **kwargs)
def resnet50(pretrained=False, **kwargs):
"""ResNet 50-layer model from
`"Deep Residual Learning for Image Recognition" <https://arxiv.org/pdf/1512.03385.pdf>`_
Args:
pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained
on ImageNet. Default: False.
Returns:
ResNet 50-layer model. An instance of :ref:`api_fluid_dygraph_Layer`.
Examples:
.. code-block:: python
import paddle
from paddle.vision.models import resnet50
# build model
model = resnet50()
# build model and load imagenet pretrained weight
# model = resnet50(pretrained=True)
x = paddle.rand([1, 3, 224, 224])
out = model(x)
print(out.shape)
# [1, 1000]
"""
return _resnet('resnet50', BottleneckBlock, 50, pretrained, **kwargs)
def resnet101(pretrained=False, **kwargs):
"""ResNet 101-layer model from
`"Deep Residual Learning for Image Recognition" <https://arxiv.org/pdf/1512.03385.pdf>`_
Args:
pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained
on ImageNet. Default: False.
Returns:
ResNet 101-layer. An instance of :ref:`api_fluid_dygraph_Layer`.
Examples:
.. code-block:: python
import paddle
from paddle.vision.models import resnet101
# build model
model = resnet101()
# build model and load imagenet pretrained weight
# model = resnet101(pretrained=True)
x = paddle.rand([1, 3, 224, 224])
out = model(x)
print(out.shape)
# [1, 1000]
"""
return _resnet('resnet101', BottleneckBlock, 101, pretrained, **kwargs)
def resnet152(pretrained=False, **kwargs):
"""ResNet 152-layer model from
`"Deep Residual Learning for Image Recognition" <https://arxiv.org/pdf/1512.03385.pdf>`_
Args:
pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained
on ImageNet. Default: False.
Returns:
ResNet 152-layer model. An instance of :ref:`api_fluid_dygraph_Layer`.
Examples:
.. code-block:: python
import paddle
from paddle.vision.models import resnet152
# build model
model = resnet152()
# build model and load imagenet pretrained weight
# model = resnet152(pretrained=True)
x = paddle.rand([1, 3, 224, 224])
out = model(x)
print(out.shape)
# [1, 1000]
"""
return _resnet('resnet152', BottleneckBlock, 152, pretrained, **kwargs)
def resnext50_32x4d(pretrained=False, **kwargs):
"""ResNeXt-50 32x4d model from
`"Aggregated Residual Transformations for Deep Neural Networks" <https://arxiv.org/pdf/1611.05431.pdf>`_
Args:
pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained
on ImageNet. Default: False.
Returns:
ResNeXt-50 32x4d model. An instance of :ref:`api_fluid_dygraph_Layer`.
Examples:
.. code-block:: python
import paddle
from paddle.vision.models import resnext50_32x4d
# build model
model = resnext50_32x4d()
# build model and load imagenet pretrained weight
# model = resnext50_32x4d(pretrained=True)
x = paddle.rand([1, 3, 224, 224])
out = model(x)
print(out.shape)
# [1, 1000]
"""
kwargs['groups'] = 32
kwargs['width'] = 4
return _resnet('resnext50_32x4d', BottleneckBlock, 50, pretrained, **kwargs)
def resnext50_64x4d(pretrained=False, **kwargs):
"""ResNeXt-50 64x4d model from
`"Aggregated Residual Transformations for Deep Neural Networks" <https://arxiv.org/pdf/1611.05431.pdf>`_
Args:
pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained
on ImageNet. Default: False.
Returns:
ResNeXt-50 64x4d model. An instance of :ref:`api_fluid_dygraph_Layer`.
Examples:
.. code-block:: python
import paddle
from paddle.vision.models import resnext50_64x4d
# build model
model = resnext50_64x4d()
# build model and load imagenet pretrained weight
# model = resnext50_64x4d(pretrained=True)
x = paddle.rand([1, 3, 224, 224])
out = model(x)
print(out.shape)
# [1, 1000]
"""
kwargs['groups'] = 64
kwargs['width'] = 4
return _resnet('resnext50_64x4d', BottleneckBlock, 50, pretrained, **kwargs)
def resnext101_32x4d(pretrained=False, **kwargs):
"""ResNeXt-101 32x4d model from
`"Aggregated Residual Transformations for Deep Neural Networks" <https://arxiv.org/pdf/1611.05431.pdf>`_
Args:
pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained
on ImageNet. Default: False.
Returns:
ResNeXt-101 32x4d model. An instance of :ref:`api_fluid_dygraph_Layer`.
Examples:
.. code-block:: python
import paddle
from paddle.vision.models import resnext101_32x4d
# build model
model = resnext101_32x4d()
# build model and load imagenet pretrained weight
# model = resnext101_32x4d(pretrained=True)
x = paddle.rand([1, 3, 224, 224])
out = model(x)
print(out.shape)
# [1, 1000]
"""
kwargs['groups'] = 32
kwargs['width'] = 4
return _resnet('resnext101_32x4d', BottleneckBlock, 101, pretrained,
**kwargs)
def resnext101_64x4d(pretrained=False, **kwargs):
"""ResNeXt-101 64x4d model from
`"Aggregated Residual Transformations for Deep Neural Networks" <https://arxiv.org/pdf/1611.05431.pdf>`_
Args:
pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained
on ImageNet. Default: False.
Returns:
ResNeXt-101 64x4d model. An instance of :ref:`api_fluid_dygraph_Layer`.
Examples:
.. code-block:: python
import paddle
from paddle.vision.models import resnext101_64x4d
# build model
model = resnext101_64x4d()
# build model and load imagenet pretrained weight
# model = resnext101_64x4d(pretrained=True)
x = paddle.rand([1, 3, 224, 224])
out = model(x)
print(out.shape)
# [1, 1000]
"""
kwargs['groups'] = 64
kwargs['width'] = 4
return _resnet('resnext101_64x4d', BottleneckBlock, 101, pretrained,
**kwargs)
def resnext152_32x4d(pretrained=False, **kwargs):
"""ResNeXt-152 32x4d model from
`"Aggregated Residual Transformations for Deep Neural Networks" <https://arxiv.org/pdf/1611.05431.pdf>`_
Args:
pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained
on ImageNet. Default: False.
Returns:
ResNeXt-152 32x4d model. An instance of :ref:`api_fluid_dygraph_Layer`.
Examples:
.. code-block:: python
import paddle
from paddle.vision.models import resnext152_32x4d
# build model
model = resnext152_32x4d()
# build model and load imagenet pretrained weight
# model = resnext152_32x4d(pretrained=True)
x = paddle.rand([1, 3, 224, 224])
out = model(x)
print(out.shape)
# [1, 1000]
"""
kwargs['groups'] = 32
kwargs['width'] = 4
return _resnet('resnext152_32x4d', BottleneckBlock, 152, pretrained,
**kwargs)
def resnext152_64x4d(pretrained=False, **kwargs):
"""ResNeXt-152 64x4d model from
`"Aggregated Residual Transformations for Deep Neural Networks" <https://arxiv.org/pdf/1611.05431.pdf>`_
Args:
pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained
on ImageNet. Default: False.
Returns:
ResNeXt-152 64x4d model. An instance of :ref:`api_fluid_dygraph_Layer`.
Examples:
.. code-block:: python
import paddle
from paddle.vision.models import resnext152_64x4d
# build model
model = resnext152_64x4d()
# build model and load imagenet pretrained weight
# model = resnext152_64x4d(pretrained=True)
x = paddle.rand([1, 3, 224, 224])
out = model(x)
print(out.shape)
# [1, 1000]
"""
kwargs['groups'] = 64
kwargs['width'] = 4
return _resnet('resnext152_64x4d', BottleneckBlock, 152, pretrained,
**kwargs)
def wide_resnet50_2(pretrained=False, **kwargs):
"""Wide ResNet-50-2 model from
`"Wide Residual Networks" <https://arxiv.org/pdf/1605.07146.pdf>`_.
Args:
pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained
on ImageNet. Default: False.
Returns:
Wide ResNet-50-2 model. An instance of :ref:`api_fluid_dygraph_Layer`.
Examples:
.. code-block:: python
import paddle
from paddle.vision.models import wide_resnet50_2
# build model
model = wide_resnet50_2()
# build model and load imagenet pretrained weight
# model = wide_resnet50_2(pretrained=True)
x = paddle.rand([1, 3, 224, 224])
out = model(x)
print(out.shape)
# [1, 1000]
"""
kwargs['width'] = 64 * 2
return _resnet('wide_resnet50_2', BottleneckBlock, 50, pretrained, **kwargs)
def wide_resnet101_2(pretrained=False, **kwargs):
"""Wide ResNet-101-2 model from
`"Wide Residual Networks" <https://arxiv.org/pdf/1605.07146.pdf>`_.
Args:
pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained
on ImageNet. Default: False.
Returns:
Wide ResNet-101-2 model. An instance of :ref:`api_fluid_dygraph_Layer`.
Examples:
.. code-block:: python
import paddle
from paddle.vision.models import wide_resnet101_2
# build model
model = wide_resnet101_2()
# build model and load imagenet pretrained weight
# model = wide_resnet101_2(pretrained=True)
x = paddle.rand([1, 3, 224, 224])
out = model(x)
print(out.shape)
# [1, 1000]
"""
kwargs['width'] = 64 * 2
return _resnet('wide_resnet101_2', BottleneckBlock, 101, pretrained,
**kwargs)