文章目录
1.加载数据
在这篇文章中，我们尝试了经典的 MNIST 数据集。使用标准卷积网络增强空间变换器网络。
from six.moves import urllib
opener = urllib.request.build_opener()
opener.addheaders = [('User-agent', 'Mozilla/5.0')]
urllib.request.install_opener(opener)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# Training dataset
train_loader = torch.utils.data.DataLoader(
datasets.MNIST(root='.', train=True, download=True,
transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])), batch_size=64, shuffle=True, num_workers=4)
# Test dataset
test_loader = torch.utils.data.DataLoader(
datasets.MNIST(root='.', train=False, transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])), batch_size=64, shuffle=True, num_workers=4)
2.什么是空间变换器网络？
空间变换器网络归结为三个主要组成部分：
- 本地网络（Localisation Network）是常规CNN，其对变换参数进行回归。不会从该数据集中明确地学习转换，而是网络自动学习增强 全局准确性的空间变换。
- 网格生成器( Grid Genator)在输入图像中生成与输出图像中的每个像素相对应的坐标网格。
- 采样器（Sampler）使用变换的参数并将其应用于输入图像。
具体该网络作用可以参考该文章
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
self.conv2_drop = nn.Dropout2d()
self.fc1 = nn.Linear(320, 50)
self.fc2 = nn.Linear(50, 10)
# Spatial transformer localization-network
self.localization = nn.Sequential(
nn.Conv2d(1, 8, kernel_size=7),
nn.MaxPool2d(2, stride=2),
nn.ReLU(True),
nn.Conv2d(8, 10, kernel_size=5),
nn.MaxPool2d(2, stride=2),
nn.ReLU(True)
)
# Regressor for the 3 * 2 affine matrix
self.fc_loc = nn.Sequential(
nn.Linear(10 * 3 * 3, 32),
nn.ReLU(True),
nn.Linear(32, 3 * 2)
)
# Initialize the weights/bias with identity transformation
self.fc_loc[2].weight.data.zero_()
self.fc_loc[2].bias.data.copy_(torch.tensor([1, 0, 0, 0, 1, 0], dtype=torch.float))
# Spatial transformer network forward function
def stn(self, x):
xs = self.localization(x)
xs = xs.view(-1, 10 * 3 * 3)
theta = self.fc_loc(xs)
theta = theta.view(-1, 2, 3)
grid = F.affine_grid(theta, x.size())
x = F.grid_sample(x, grid)
return x
def forward(self, x):
# transform the input
x = self.stn(x)
# Perform the usual forward pass
x = F.relu(F.max_pool2d(self.conv1(x), 2))
x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2))
x = x.view(-1, 320)
x = F.relu(self.fc1(x))
x = F.dropout(x, training=self.training)
x = self.fc2(x)
return F.log_softmax(x, dim=1)
model = Net().to(device)
3.训练模型
训练模型 现在我们使用 SGD（随机梯度下降）算法来训练模型。网络正在以有监督的方式学习分
类任务。同时，该模型以端到端的方式自动学习STN。
optimizer = optim.SGD(model.parameters(), lr=0.01)
def train(epoch):
model.train()
for batch_idx, (data, target) in enumerate(train_loader):
data, target = data.to(device), target.to(device)
optimizer.zero_grad()
output = model(data)
loss = F.nll_loss(output, target)
loss.backward()
optimizer.step()
if batch_idx % 500 == 0:
print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
epoch, batch_idx * len(data), len(train_loader.dataset),
100. * batch_idx / len(train_loader), loss.item()))
#
# A simple test procedure to measure the STN performances on MNIST.
#
def test():
with torch.no_grad():
model.eval()
test_loss = 0
correct = 0
for data, target in test_loader:
data, target = data.to(device), target.to(device)
output = model(data)
# sum up batch loss
test_loss += F.nll_loss(output, target, size_average=False).item()
# get the index of the max log-probability
pred = output.max(1, keepdim=True)[1]
correct += pred.eq(target.view_as(pred)).sum().item()
test_loss /= len(test_loader.dataset)
print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'
.format(test_loss, correct, len(test_loader.dataset),
100. * correct / len(test_loader.dataset)))
4.可视化 STN 结果
现在，我们将检查我们学习的视觉注意机制的结果。
我们定义了一个小辅助函数，以便在训练时可视化变换。
def convert_image_np(inp):
"""Convert a Tensor to numpy image."""
inp = inp.numpy().transpose((1, 2, 0))
mean = np.array([0.485, 0.456, 0.406])
std = np.array([0.229, 0.224, 0.225])
inp = std * inp + mean
inp = np.clip(inp, 0, 1)
return inp
# We want to visualize the output of the spatial transformers layer
# after the training, we visualize a batch of input images and
# the corresponding transformed batch using STN.
def visualize_stn():
with torch.no_grad():
# Get a batch of training data
data = next(iter(test_loader))[0].to(device)
input_tensor = data.cpu()
transformed_input_tensor = model.stn(data).cpu()
in_grid = convert_image_np(
torchvision.utils.make_grid(input_tensor))
out_grid = convert_image_np(
torchvision.utils.make_grid(transformed_input_tensor))
# Plot the results side-by-side
f, axarr = plt.subplots(1, 2)
axarr[0].imshow(in_grid)
axarr[0].set_title('Dataset Images')
axarr[1].imshow(out_grid)
axarr[1].set_title('Transformed Images')
if __name__ == '__main__':
for epoch in range(1, 20 + 1):
train(epoch)
test()
# Visualize the STN transformation on some input batch
visualize_stn()
plt.ioff()
plt.show()
输出结果：
Train Epoch: 1 [0/60000 (0%)] Loss: 2.317233
Train Epoch: 1 [32000/60000 (53%)] Loss: 1.035397
Test set: Average loss: 0.2624, Accuracy: 9272/10000 (93%)
Train Epoch: 2 [0/60000 (0%)] Loss: 0.524522
Train Epoch: 2 [32000/60000 (53%)] Loss: 0.519621
Test set: Average loss: 0.2160, Accuracy: 9328/10000 (93%)
Train Epoch: 3 [0/60000 (0%)] Loss: 0.677090
Train Epoch: 3 [32000/60000 (53%)] Loss: 0.174913
Test set: Average loss: 0.0999, Accuracy: 9676/10000 (97%)
Train Epoch: 4 [0/60000 (0%)] Loss: 0.280676
Train Epoch: 4 [32000/60000 (53%)] Loss: 0.448485
Test set: Average loss: 0.0800, Accuracy: 9755/10000 (98%)
Train Epoch: 5 [0/60000 (0%)] Loss: 0.364630
Train Epoch: 5 [32000/60000 (53%)] Loss: 0.380116
Test set: Average loss: 0.0710, Accuracy: 9785/10000 (98%)
Train Epoch: 6 [0/60000 (0%)] Loss: 0.233124
Train Epoch: 6 [32000/60000 (53%)] Loss: 0.080830
Test set: Average loss: 0.0661, Accuracy: 9798/10000 (98%)
Train Epoch: 7 [0/60000 (0%)] Loss: 0.093886
Train Epoch: 7 [32000/60000 (53%)] Loss: 0.117194
Test set: Average loss: 0.0621, Accuracy: 9821/10000 (98%)
Train Epoch: 8 [0/60000 (0%)] Loss: 0.164851
Train Epoch: 8 [32000/60000 (53%)] Loss: 0.084818
Test set: Average loss: 0.1057, Accuracy: 9682/10000 (97%)
Train Epoch: 9 [0/60000 (0%)] Loss: 0.178490
Train Epoch: 9 [32000/60000 (53%)] Loss: 0.066079
Test set: Average loss: 0.0614, Accuracy: 9822/10000 (98%)
Train Epoch: 10 [0/60000 (0%)] Loss: 0.090977
Train Epoch: 10 [32000/60000 (53%)] Loss: 0.106222
Test set: Average loss: 0.0610, Accuracy: 9813/10000 (98%)
Train Epoch: 11 [0/60000 (0%)] Loss: 0.079447
Train Epoch: 11 [32000/60000 (53%)] Loss: 0.103146
Test set: Average loss: 0.0516, Accuracy: 9851/10000 (99%)
Train Epoch: 12 [0/60000 (0%)] Loss: 0.096039
Train Epoch: 12 [32000/60000 (53%)] Loss: 0.152670
Test set: Average loss: 0.0871, Accuracy: 9754/10000 (98%)
Train Epoch: 13 [0/60000 (0%)] Loss: 0.250616
Train Epoch: 13 [32000/60000 (53%)] Loss: 0.095682
Test set: Average loss: 0.0510, Accuracy: 9855/10000 (99%)
Train Epoch: 14 [0/60000 (0%)] Loss: 0.159322
Train Epoch: 14 [32000/60000 (53%)] Loss: 0.093004
Test set: Average loss: 0.0486, Accuracy: 9856/10000 (99%)
Train Epoch: 15 [0/60000 (0%)] Loss: 0.273801
Train Epoch: 15 [32000/60000 (53%)] Loss: 0.220219
Test set: Average loss: 0.0568, Accuracy: 9829/10000 (98%)
Train Epoch: 16 [0/60000 (0%)] Loss: 0.237934
Train Epoch: 16 [32000/60000 (53%)] Loss: 0.105854
Test set: Average loss: 0.0538, Accuracy: 9846/10000 (98%)
Train Epoch: 17 [0/60000 (0%)] Loss: 0.409603
Train Epoch: 17 [32000/60000 (53%)] Loss: 0.110827
Test set: Average loss: 0.0843, Accuracy: 9753/10000 (98%)
Train Epoch: 18 [0/60000 (0%)] Loss: 0.115920
Train Epoch: 18 [32000/60000 (53%)] Loss: 0.143565
Test set: Average loss: 0.0404, Accuracy: 9883/10000 (99%)
Train Epoch: 19 [0/60000 (0%)] Loss: 0.027614
Train Epoch: 19 [32000/60000 (53%)] Loss: 0.086963
Test set: Average loss: 0.0526, Accuracy: 9849/10000 (98%)
Train Epoch: 20 [0/60000 (0%)] Loss: 0.229980
Train Epoch: 20 [32000/60000 (53%)] Loss: 0.210401
Test set: Average loss: 0.0462, Accuracy: 9859/10000 (99%)