（三）Pytorch常用迁移学习深度模型结构

最新推荐文章于 2024-10-07 18:53:43 发布

Huaping Guo

最新推荐文章于 2024-10-07 18:53:43 发布

阅读量661

点赞数

文章标签：深度学习

本文链接：https://blog.csdn.net/hpguo_cm/article/details/113062212

版权

SqueezeNet

压缩策略

将 $3 * 3$ 卷积替换成 $1 * 1$ 卷积：通过这一步，一个卷积操作的参数数量减少了9倍；
减少 $3 * 3$ 卷积的通道数：一个 $3 * 3$ 卷积的计算量是 $3 * 3 * M * N$ （其中 $M$ ， $N$ 分别是输入Feature Map和输出Feature Map的通道数），作者降低 $M$ ， $N$ 值以减少参数数量；
将降采样后置：作者认为较大的Feature Map含有更多的信息，因此将降采样往分类层移动。

总体结构

pytorch自带SqueezeNet，使用下面代码可以查看SqueezeNet结构

代码

from torchvision import models
mod = models.SqueezeNet(version='1_0')#默认为1.0版本
print(mod)

1.0版本结构

SqueezeNet(
  (features): Sequential(
    (0): Conv2d(3, 96, kernel_size=(7, 7), stride=(2, 2))
    (1): ReLU(inplace)
    (2): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=True)
    (3): Fire(
      (squeeze): Conv2d(96, 16, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(16, 64, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(16, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
    (4): Fire(
      (squeeze): Conv2d(128, 16, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(16, 64, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(16, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
    (5): Fire(
      (squeeze): Conv2d(128, 32, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(32, 128, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(32, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
    (6): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=True)
    (7): Fire(
      (squeeze): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(32, 128, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(32, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
    (8): Fire(
      (squeeze): Conv2d(256, 48, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(48, 192, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(48, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
    (9): Fire(
      (squeeze): Conv2d(384, 48, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(48, 192, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(48, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
    (10): Fire(
      (squeeze): Conv2d(384, 64, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(64, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
    (11): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=True)
    (12): Fire(
      (squeeze): Conv2d(512, 64, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(64, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
  )
  (classifier): Sequential(
    (0): Dropout(p=0.5)
    (1): Conv2d(512, 1000, kernel_size=(1, 1), stride=(1, 1))
    (2): ReLU(inplace)
    (3): AdaptiveAvgPool2d(output_size=(1, 1))
  )
)

Fire结构
在SqueezeNet总体结构中，Fire部件是最重要的的组成部件。一个Fire模块由Squeeze部分和Expand部分组成。具体结构如下，要求fire2_e1+fire2_e3 > fire2_s1
1.1版本结构

使用如下代码查看SqueezeNet结构

from torchvision import models
mod = models.SqueezeNet(version='2_0')#默认为1.0版本
print(mod)

结构如下：

SqueezeNet(
  (features): Sequential(
    (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(2, 2))
    (1): ReLU(inplace)
    (2): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=True)
    (3): Fire(
      (squeeze): Conv2d(64, 16, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(16, 64, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(16, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
    (4): Fire(
      (squeeze): Conv2d(128, 16, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(16, 64, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(16, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
    (5): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=True)
    (6): Fire(
      (squeeze): Conv2d(128, 32, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(32, 128, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(32, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
    (7): Fire(
      (squeeze): Conv2d(256, 32, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(32, 128, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(32, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
    (8): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=True)
    (9): Fire(
      (squeeze): Conv2d(256, 48, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(48, 192, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(48, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
    (10): Fire(
      (squeeze): Conv2d(384, 48, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(48, 192, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(48, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
    (11): Fire(
      (squeeze): Conv2d(384, 64, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(64, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
    (12): Fire(
      (squeeze): Conv2d(512, 64, kernel_size=(1, 1), stride=(1, 1))
      (squeeze_activation): ReLU(inplace)
      (expand1x1): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))
      (expand1x1_activation): ReLU(inplace)
      (expand3x3): Conv2d(64, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (expand3x3_activation): ReLU(inplace)
    )
  )
  (classifier): Sequential(
    (0): Dropout(p=0.5)
    (1): Conv2d(512, 1000, kernel_size=(1, 1), stride=(1, 1))
    (2): ReLU(inplace)
    (3): AdaptiveAvgPool2d(output_size=(1, 1))
  )
)

参考文献

F. N. Iandola, S. Han, M. W. Moskewicz, K. Ashraf, W. J. Dally, and K. Keutzer, SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5 MB model size, 2016, arXiv:1602.07360. pdf

VGG16

模型特点

1.结构单一，所有的卷积核尺寸都是3x3，池化核尺寸都是2x2，所有的激活函数都是Relu；
2. 计算量小，采用连续的几个3x3的卷积核代替AlexNet中的较大卷积核（11x11，7x7，5x5），保证了感受野不变的情况下，减少了计算量；

总体结构

pytorch自带VGG16，使用下面代码可以查看VGG结构

代码

from torchvision import models
mod = models.vgg16(pretrained=True)
print(mod)

结构图
结构

VGG(
  (features): Sequential(
    (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): ReLU(inplace)
    (2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (3): ReLU(inplace)
    (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (5): Conv2d