空间变压器网络教程

原文： https://pytorch.org/tutorials/intermediate/spatial_transformer_tutorial.html

注意

单击此处的下载完整的示例代码

作者： Ghassen HAMROUNI

../_images/FSeq.png

在本教程中，您将学习如何使用称为空间变换器网络的视觉注意力机制来扩充网络。 您可以在 DeepMind 论文中详细了解空间变压器网络。

空间变换器网络是对任何空间变换的可区别关注的概括。 空间变换器网络(简称 STN）允许神经网络学习如何对输入图像执行空间变换，以增强模型的几何不变性。 例如，它可以裁剪感兴趣的区域，缩放并校正图像的方向。 这可能是一个有用的机制，因为 CNN 不会对旋转和缩放以及更一般的仿射变换保持不变。

关于 STN 的最好的事情之一就是能够将它简单地插入到任何现有的 CNN 中。

  1. # License: BSD
  2. # Author: Ghassen Hamrouni
  4. from __future__ import print_function
  5. import torch
  6. import torch.nn as nn
  7. import torch.nn.functional as F
  8. import torch.optim as optim
  9. import torchvision
  10. from torchvision import datasets, transforms
  11. import matplotlib.pyplot as plt
  12. import numpy as np
  14. plt.ion()   # interactive mode

加载数据

在本文中，我们将尝试使用经典的 MNIST 数据集。 使用标准卷积网络和空间变换器网络。

  1. device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
  3. # Training dataset
  4. train_loader = torch.utils.data.DataLoader(
  5.     datasets.MNIST(root='.', train=True, download=True,
  6.                    transform=transforms.Compose([
  7.                        transforms.ToTensor(),
  8.                        transforms.Normalize((0.1307,), (0.3081,))
  9.                    ])), batch_size=64, shuffle=True, num_workers=4)
  10. # Test dataset
  11. test_loader = torch.utils.data.DataLoader(
  12.     datasets.MNIST(root='.', train=False, transform=transforms.Compose([
  13.         transforms.ToTensor(),
  14.         transforms.Normalize((0.1307,), (0.3081,))
  15.     ])), batch_size=64, shuffle=True, num_workers=4)

出：

  1. Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz to ./MNIST/raw/train-images-idx3-ubyte.gz
  2. Extracting ./MNIST/raw/train-images-idx3-ubyte.gz to ./MNIST/raw
  3. Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz to ./MNIST/raw/train-labels-idx1-ubyte.gz
  4. Extracting ./MNIST/raw/train-labels-idx1-ubyte.gz to ./MNIST/raw
  5. Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz to ./MNIST/raw/t10k-images-idx3-ubyte.gz
  6. Extracting ./MNIST/raw/t10k-images-idx3-ubyte.gz to ./MNIST/raw
  7. Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz to ./MNIST/raw/t10k-labels-idx1-ubyte.gz
  8. Extracting ./MNIST/raw/t10k-labels-idx1-ubyte.gz to ./MNIST/raw
  9. Processing...
  10. Done!

描述空间变压器网络

空间变压器网络可归结为三个主要组成部分：

  • 本地化网络是常规的 CNN，可以对转换参数进行回归。 永远不会从此数据集中显式学习变换，而是网络会自动学习增强全局精度的空间变换。
  • 网格生成器在输入图像中生成与来自输出图像的每个像素相对应的坐标网格。
  • 采样器使用转换的参数，并将其应用于输入图像。

../_images/stn-arch.png

Note

我们需要包含 affine_grid 和 grid_sample 模块的最新版本的 PyTorch。

  1. class Net(nn.Module):
  2.     def __init__(self):
  3.         super(Net, self).__init__()
  4.         self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
  5.         self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
  6.         self.conv2_drop = nn.Dropout2d()
  7.         self.fc1 = nn.Linear(320, 50)
  8.         self.fc2 = nn.Linear(50, 10)
  10.         # Spatial transformer localization-network
  11.         self.localization = nn.Sequential(
  12.             nn.Conv2d(1, 8, kernel_size=7),
  13.             nn.MaxPool2d(2, stride=2),
  14.             nn.ReLU(True),
  15.             nn.Conv2d(8, 10, kernel_size=5),
  16.             nn.MaxPool2d(2, stride=2),
  17.             nn.ReLU(True)
  18.         )
  20.         # Regressor for the 3 * 2 affine matrix
  21.         self.fc_loc = nn.Sequential(
  22.             nn.Linear(10 * 3 * 3, 32),
  23.             nn.ReLU(True),
  24.             nn.Linear(32, 3 * 2)
  25.         )
  27.         # Initialize the weights/bias with identity transformation
  28.         self.fc_loc[2].weight.data.zero_()
  29.         self.fc_loc[2].bias.data.copy_(torch.tensor([1, 0, 0, 0, 1, 0], dtype=torch.float))
  31.     # Spatial transformer network forward function
  32.     def stn(self, x):
  33.         xs = self.localization(x)
  34.         xs = xs.view(-1, 10 * 3 * 3)
  35.         theta = self.fc_loc(xs)
  36.         theta = theta.view(-1, 2, 3)
  38.         grid = F.affine_grid(theta, x.size())
  39.         x = F.grid_sample(x, grid)
  41.         return x
  43.     def forward(self, x):
  44.         # transform the input
  45.         x = self.stn(x)
  47.         # Perform the usual forward pass
  48.         x = F.relu(F.max_pool2d(self.conv1(x), 2))
  49.         x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2))
  50.         x = x.view(-1, 320)
  51.         x = F.relu(self.fc1(x))
  52.         x = F.dropout(x, training=self.training)
  53.         x = self.fc2(x)
  54.         return F.log_softmax(x, dim=1)
  56. model = Net().to(device)

训练模型

现在，让我们使用 SGD 算法训练模型。 网络正在以监督方式学习分类任务。 同时，该模型以端到端的方式自动学习 STN。

  1. optimizer = optim.SGD(model.parameters(), lr=0.01)
  3. def train(epoch):
  4.     model.train()
  5.     for batch_idx, (data, target) in enumerate(train_loader):
  6.         data, target = data.to(device), target.to(device)
  8.         optimizer.zero_grad()
  9.         output = model(data)
  10.         loss = F.nll_loss(output, target)
  11.         loss.backward()
  12.         optimizer.step()
  13.         if batch_idx % 500 == 0:
  14.             print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
  15.                 epoch, batch_idx * len(data), len(train_loader.dataset),
  16.                 100\. * batch_idx / len(train_loader), loss.item()))
  17. #
  18. # A simple test procedure to measure STN the performances on MNIST.
  19. #
  21. def test():
  22.     with torch.no_grad():
  23.         model.eval()
  24.         test_loss = 0
  25.         correct = 0
  26.         for data, target in test_loader:
  27.             data, target = data.to(device), target.to(device)
  28.             output = model(data)
  30.             # sum up batch loss
  31.             test_loss += F.nll_loss(output, target, size_average=False).item()
  32.             # get the index of the max log-probability
  33.             pred = output.max(1, keepdim=True)[1]
  34.             correct += pred.eq(target.view_as(pred)).sum().item()
  36.         test_loss /= len(test_loader.dataset)
  37.         print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'
  38.               .format(test_loss, correct, len(test_loader.dataset),
  39.                       100\. * correct / len(test_loader.dataset)))

可视化 STN 结果

现在，我们将检查学习到的视觉注意力机制的结果。

我们定义了一个小的辅助函数，以便在训练时可视化转换。

  1. def convert_image_np(inp):
  2.     """Convert a Tensor to numpy image."""
  3.     inp = inp.numpy().transpose((1, 2, 0))
  4.     mean = np.array([0.485, 0.456, 0.406])
  5.     std = np.array([0.229, 0.224, 0.225])
  6.     inp = std * inp + mean
  7.     inp = np.clip(inp, 0, 1)
  8.     return inp
  10. # We want to visualize the output of the spatial transformers layer
  11. # after the training, we visualize a batch of input images and
  12. # the corresponding transformed batch using STN.
  14. def visualize_stn():
  15.     with torch.no_grad():
  16.         # Get a batch of training data
  17.         data = next(iter(test_loader))[0].to(device)
  19.         input_tensor = data.cpu()
  20.         transformed_input_tensor = model.stn(data).cpu()
  22.         in_grid = convert_image_np(
  23.             torchvision.utils.make_grid(input_tensor))
  25.         out_grid = convert_image_np(
  26.             torchvision.utils.make_grid(transformed_input_tensor))
  28.         # Plot the results side-by-side
  29.         f, axarr = plt.subplots(1, 2)
  30.         axarr[0].imshow(in_grid)
  31.         axarr[0].set_title('Dataset Images')
  33.         axarr[1].imshow(out_grid)
  34.         axarr[1].set_title('Transformed Images')
  36. for epoch in range(1, 20 + 1):
  37.     train(epoch)
  38.     test()
  40. # Visualize the STN transformation on some input batch
  41. visualize_stn()
  43. plt.ioff()
  44. plt.show()

../_images/sphx_glr_spatial_transformer_tutorial_001.png

Out:

  1. Train Epoch: 1 [0/60000 (0%)]   Loss: 2.312544
  2. Train Epoch: 1 [32000/60000 (53%)]      Loss: 0.865688
  4. Test set: Average loss: 0.2105, Accuracy: 9426/10000 (94%)
  6. Train Epoch: 2 [0/60000 (0%)]   Loss: 0.528199
  7. Train Epoch: 2 [32000/60000 (53%)]      Loss: 0.273284
  9. Test set: Average loss: 0.1150, Accuracy: 9661/10000 (97%)
  11. Train Epoch: 3 [0/60000 (0%)]   Loss: 0.312562
  12. Train Epoch: 3 [32000/60000 (53%)]      Loss: 0.496166
  14. Test set: Average loss: 0.1130, Accuracy: 9661/10000 (97%)
  16. Train Epoch: 4 [0/60000 (0%)]   Loss: 0.346181
  17. Train Epoch: 4 [32000/60000 (53%)]      Loss: 0.206084
  19. Test set: Average loss: 0.0875, Accuracy: 9730/10000 (97%)
  21. Train Epoch: 5 [0/60000 (0%)]   Loss: 0.351175
  22. Train Epoch: 5 [32000/60000 (53%)]      Loss: 0.388225
  24. Test set: Average loss: 0.0659, Accuracy: 9802/10000 (98%)
  26. Train Epoch: 6 [0/60000 (0%)]   Loss: 0.122667
  27. Train Epoch: 6 [32000/60000 (53%)]      Loss: 0.258372
  29. Test set: Average loss: 0.0791, Accuracy: 9759/10000 (98%)
  31. Train Epoch: 7 [0/60000 (0%)]   Loss: 0.190197
  32. Train Epoch: 7 [32000/60000 (53%)]      Loss: 0.154468
  34. Test set: Average loss: 0.0647, Accuracy: 9791/10000 (98%)
  36. Train Epoch: 8 [0/60000 (0%)]   Loss: 0.121149
  37. Train Epoch: 8 [32000/60000 (53%)]      Loss: 0.288490
  39. Test set: Average loss: 0.0583, Accuracy: 9821/10000 (98%)
  41. Train Epoch: 9 [0/60000 (0%)]   Loss: 0.244609
  42. Train Epoch: 9 [32000/60000 (53%)]      Loss: 0.023396
  44. Test set: Average loss: 0.0685, Accuracy: 9778/10000 (98%)
  46. Train Epoch: 10 [0/60000 (0%)]  Loss: 0.256878
  47. Train Epoch: 10 [32000/60000 (53%)]     Loss: 0.091626
  49. Test set: Average loss: 0.0684, Accuracy: 9783/10000 (98%)
  51. Train Epoch: 11 [0/60000 (0%)]  Loss: 0.181910
  52. Train Epoch: 11 [32000/60000 (53%)]     Loss: 0.113193
  54. Test set: Average loss: 0.0492, Accuracy: 9856/10000 (99%)
  56. Train Epoch: 12 [0/60000 (0%)]  Loss: 0.081072
  57. Train Epoch: 12 [32000/60000 (53%)]     Loss: 0.082513
  59. Test set: Average loss: 0.0670, Accuracy: 9800/10000 (98%)
  61. Train Epoch: 13 [0/60000 (0%)]  Loss: 0.180748
  62. Train Epoch: 13 [32000/60000 (53%)]     Loss: 0.194512
  64. Test set: Average loss: 0.0439, Accuracy: 9874/10000 (99%)
  66. Train Epoch: 14 [0/60000 (0%)]  Loss: 0.099560
  67. Train Epoch: 14 [32000/60000 (53%)]     Loss: 0.084377
  69. Test set: Average loss: 0.0416, Accuracy: 9880/10000 (99%)
  71. Train Epoch: 15 [0/60000 (0%)]  Loss: 0.070021
  72. Train Epoch: 15 [32000/60000 (53%)]     Loss: 0.241336
  74. Test set: Average loss: 0.0588, Accuracy: 9820/10000 (98%)
  76. Train Epoch: 16 [0/60000 (0%)]  Loss: 0.060536
  77. Train Epoch: 16 [32000/60000 (53%)]     Loss: 0.053016
  79. Test set: Average loss: 0.0405, Accuracy: 9877/10000 (99%)
  81. Train Epoch: 17 [0/60000 (0%)]  Loss: 0.207369
  82. Train Epoch: 17 [32000/60000 (53%)]     Loss: 0.069607
  84. Test set: Average loss: 0.1006, Accuracy: 9685/10000 (97%)
  86. Train Epoch: 18 [0/60000 (0%)]  Loss: 0.127503
  87. Train Epoch: 18 [32000/60000 (53%)]     Loss: 0.070724
  89. Test set: Average loss: 0.0659, Accuracy: 9814/10000 (98%)
  91. Train Epoch: 19 [0/60000 (0%)]  Loss: 0.176861
  92. Train Epoch: 19 [32000/60000 (53%)]     Loss: 0.116980
  94. Test set: Average loss: 0.0413, Accuracy: 9871/10000 (99%)
  96. Train Epoch: 20 [0/60000 (0%)]  Loss: 0.146933
  97. Train Epoch: 20 [32000/60000 (53%)]     Loss: 0.245741
  99. Test set: Average loss: 0.0346, Accuracy: 9892/10000 (99%)

脚本的总运行时间：(2 分钟 3.339 秒）

Download Python source code: spatial_transformer_tutorial.py Download Jupyter notebook: spatial_transformer_tutorial.ipynb

由狮身人面像画廊生成的画廊