学习表征

在 2012 年前,图像的特征工程是关键,SIFT、SURF、HOG(定向梯度直方图)等特征提取方法占据主导地位

2009 年,ImageNet 数据集发布,图片是自然物体的彩色图片,图片大小为 469×387,样本数为 1.2M,类数为 1000

2021 年 Alex Krizhevsky 提出的一种新的卷积神经网络变体 AlexNet 赢得了 2012 年 ImageNet 竞赛

深度卷积神经网络的突破可归因于数据和硬件两个因素

Untitled

AlexNet

更深更大的 LeNet

主要改进:Dropout、ReLU、MaxPooling

引领计算机视觉方法论的改变:从“人工特征提取→SVM”到“通过 CNN 学习特征→Softmax 回归”

Untitled

AlexNet 由八层组成:五个卷积层、两个全连接隐藏层和一个全连接输出层。

模型设计

在 AlexNet 的第一层,卷积窗口的形状是11×11。由于 ImageNet 中大多数图像的宽和高比 MNIST 图像的多 10 倍以上,因此需要一个更大的卷积窗口来捕获目标。

第二层中的卷积窗口形状被缩减为 5×5,然后是 3×3。

在第一层、第二层和第五层卷积层之后,加入窗口形状为 3×3、步幅为 2 的最大池化层。

在最后一个卷积层后有两个全连接层,分别有 4096 个输出。 这两个巨大的全连接层拥有将近 1GB 的模型参数。

更多细节

AlexNet 使用 ReLU 而不是 sigmoid 作为其激活函数,可以减缓梯度消失。

隐藏全连接层后加入了 Dropout 层来控制全连接层的模型复杂度。

为了进一步扩充数据,AlexNet 在训练时增加了大量的图像增强数据,如翻转、裁切和变色,使得模型更健壮,更大的样本量有效地减少了过拟合

Untitled

总结

  • AlexNet 的架构与 LeNet 相似,但使用了更多的卷积层和更多的参数来拟合大规模的ImageNet数据集,达到了 大约10 倍的参数量和 260 倍的计算量。
  • 新引入了 Dropout、ReLU、最大池化层和数据增强
  • AlexNet 标志着新一轮神经网络热潮的开始
  • Dropout、ReLU 和预处理是提升计算机视觉任务性能的其他关键步骤

代码实现

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
import torch
from torch import nn
from d2l import torch as d2l

net = nn.Sequential(
# 这里,我们使用一个11*11的更大窗口来捕捉对象。
# 同时,步幅为4,以减少输出的高度和宽度。
# 另外,输出通道的数目远大于LeNet
nn.Conv2d(1, 96, kernel_size=11, stride=4, padding=1), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
# 减小卷积窗口,使用填充为2来使得输入与输出的高和宽一致,且增大输出通道数
nn.Conv2d(96, 256, kernel_size=5, padding=2), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
# 使用三个连续的卷积层和较小的卷积窗口。
# 除了最后的卷积层,输出通道的数量进一步增加。
# 在前两个卷积层之后,汇聚层不用于减少输入的高度和宽度
nn.Conv2d(256, 384, kernel_size=3, padding=1), nn.ReLU(),
nn.Conv2d(384, 384, kernel_size=3, padding=1), nn.ReLU(),
nn.Conv2d(384, 256, kernel_size=3, padding=1), nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Flatten(),
# 这里,全连接层的输出数量是LeNet中的好几倍。使用dropout层来减轻过拟合
nn.Linear(6400, 4096), nn.ReLU(),
nn.Dropout(p=0.5),
nn.Linear(4096, 4096), nn.ReLU(),
nn.Dropout(p=0.5),
# 最后是输出层。由于这里使用Fashion-MNIST,所以用类别数为10,而非论文中的1000
nn.Linear(4096, 10))

构造一个高度和宽度都为 224 的单通道数据,来观察每一层输出的形状:

1
2
3
4
X = torch.randn(1, 1, 224, 224)
for layer in net:
X=layer(X)
print(layer.__class__.__name__,'output shape:\t',X.shape)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Conv2d output shape:         torch.Size([1, 96, 54, 54])
ReLU output shape: torch.Size([1, 96, 54, 54])
MaxPool2d output shape: torch.Size([1, 96, 26, 26])
Conv2d output shape: torch.Size([1, 256, 26, 26])
ReLU output shape: torch.Size([1, 256, 26, 26])
MaxPool2d output shape: torch.Size([1, 256, 12, 12])
Conv2d output shape: torch.Size([1, 384, 12, 12])
ReLU output shape: torch.Size([1, 384, 12, 12])
Conv2d output shape: torch.Size([1, 384, 12, 12])
ReLU output shape: torch.Size([1, 384, 12, 12])
Conv2d output shape: torch.Size([1, 256, 12, 12])
ReLU output shape: torch.Size([1, 256, 12, 12])
MaxPool2d output shape: torch.Size([1, 256, 5, 5])
Flatten output shape: torch.Size([1, 6400])
Linear output shape: torch.Size([1, 4096])
ReLU output shape: torch.Size([1, 4096])
Dropout output shape: torch.Size([1, 4096])
Linear output shape: torch.Size([1, 4096])
ReLU output shape: torch.Size([1, 4096])
Dropout output shape: torch.Size([1, 4096])
Linear output shape: torch.Size([1, 10])

Fashion-MNIST 图像的分辨率(28×28像素)低于 ImageNet 图像,将它们增加到 224×224(只是为了展示 AlexNet 的架构),使用d2l.load_data_fashion_mnist函数中的resize参数执行此调整:

1
2
batch_size = 128
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size, resize=224)

训练 AlexNet:

1
2
lr, num_epochs = 0.01, 10
d2l.train_ch6(net, train_iter, test_iter, num_epochs, lr, d2l.try_gpu())
1
2
loss 0.333, train acc 0.877, test acc 0.879
4152.2 examples/sec on cuda:0

Untitled