Spatial Transformer Networks Tutorial

Created On: Nov 08, 2017 | Last Updated: Jan 19, 2024 | Last Verified: Nov 05, 2024

Author: Ghassen HAMROUNI

In this tutorial, you will learn how to augment your network using a visual attention mechanism called spatial transformer networks. You can read more about the spatial transformer networks in the DeepMind paper

Spatial transformer networks are a generalization of differentiable attention to any spatial transformation. Spatial transformer networks (STN for short) allow a neural network to learn how to perform spatial transformations on the input image in order to enhance the geometric invariance of the model. For example, it can crop a region of interest, scale and correct the orientation of an image. It can be a useful mechanism because CNNs are not invariant to rotation and scale and more general affine transformations.

One of the best things about STN is the ability to simply plug it into any existing CNN with very little modification.

# License: BSD
# Author: Ghassen Hamrouni

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import torchvision
from torchvision import datasets, transforms
import matplotlib.pyplot as plt
import numpy as np

plt.ion()   # interactive mode

<contextlib.ExitStack object at 0x7f40c2824550>

Loading the data

In this post we experiment with the classic MNIST dataset. Using a standard convolutional network augmented with a spatial transformer network.

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)

  0%|          | 0.00/9.91M [00:00<?, ?B/s]
100%|##########| 9.91M/9.91M [00:00<00:00, 137MB/s]

  0%|          | 0.00/28.9k [00:00<?, ?B/s]
100%|##########| 28.9k/28.9k [00:00<00:00, 36.4MB/s]

  0%|          | 0.00/1.65M [00:00<?, ?B/s]
100%|##########| 1.65M/1.65M [00:00<00:00, 291MB/s]

  0%|          | 0.00/4.54k [00:00<?, ?B/s]
100%|##########| 4.54k/4.54k [00:00<00:00, 30.4MB/s]

Depicting spatial transformer networks

Spatial transformer networks boils down to three main components :

• The localization network is a regular CNN which regresses the transformation parameters. The transformation is never learned explicitly from this dataset, instead the network learns automatically the spatial transformations that enhances the global accuracy.

• The grid generator generates a grid of coordinates in the input image corresponding to each pixel from the output image.

• The sampler uses the parameters of the transformation and applies it to the input image.

Note

We need the latest version of PyTorch that contains affine_grid and grid_sample modules.

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)

Training the model

Now, let’s use the SGD algorithm to train the model. The network is learning the classification task in a supervised way. In the same time the model is learning STN automatically in an end-to-end fashion.

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)))

Visualizing the STN results

Now, we will inspect the results of our learned visual attention mechanism.

We define a small helper function in order to visualize the transformations while training.

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')

for epoch in range(1, 20 + 1):
    train(epoch)
    test()

# Visualize the STN transformation on some input batch
visualize_stn()

plt.ioff()
plt.show()

/usr/local/lib/python3.10/dist-packages/torch/nn/functional.py:5082: UserWarning:

Default grid_sample and affine_grid behavior has changed to align_corners=False since 1.3.0. Please specify align_corners=True if the old behavior is desired. See the documentation of grid_sample for details.

/usr/local/lib/python3.10/dist-packages/torch/nn/functional.py:5015: UserWarning:

Default grid_sample and affine_grid behavior has changed to align_corners=False since 1.3.0. Please specify align_corners=True if the old behavior is desired. See the documentation of grid_sample for details.

Train Epoch: 1 [0/60000 (0%)]   Loss: 2.326503
Train Epoch: 1 [32000/60000 (53%)]      Loss: 0.874084
/usr/local/lib/python3.10/dist-packages/torch/nn/_reduction.py:51: UserWarning:

size_average and reduce args will be deprecated, please use reduction='sum' instead.


Test set: Average loss: 0.2162, Accuracy: 9381/10000 (94%)

Train Epoch: 2 [0/60000 (0%)]   Loss: 0.626854
Train Epoch: 2 [32000/60000 (53%)]      Loss: 0.268811

Test set: Average loss: 0.1076, Accuracy: 9677/10000 (97%)

Train Epoch: 3 [0/60000 (0%)]   Loss: 0.278225
Train Epoch: 3 [32000/60000 (53%)]      Loss: 0.184019

Test set: Average loss: 0.0844, Accuracy: 9738/10000 (97%)

Train Epoch: 4 [0/60000 (0%)]   Loss: 0.111223
Train Epoch: 4 [32000/60000 (53%)]      Loss: 0.195178

Test set: Average loss: 0.0994, Accuracy: 9695/10000 (97%)

Train Epoch: 5 [0/60000 (0%)]   Loss: 0.104982
Train Epoch: 5 [32000/60000 (53%)]      Loss: 0.307171

Test set: Average loss: 0.5644, Accuracy: 8298/10000 (83%)

Train Epoch: 6 [0/60000 (0%)]   Loss: 1.205783
Train Epoch: 6 [32000/60000 (53%)]      Loss: 0.526758

Test set: Average loss: 0.0584, Accuracy: 9829/10000 (98%)

Train Epoch: 7 [0/60000 (0%)]   Loss: 0.129269
Train Epoch: 7 [32000/60000 (53%)]      Loss: 0.129555

Test set: Average loss: 0.1102, Accuracy: 9638/10000 (96%)

Train Epoch: 8 [0/60000 (0%)]   Loss: 0.214305
Train Epoch: 8 [32000/60000 (53%)]      Loss: 0.173208

Test set: Average loss: 0.0713, Accuracy: 9772/10000 (98%)

Train Epoch: 9 [0/60000 (0%)]   Loss: 0.163014
Train Epoch: 9 [32000/60000 (53%)]      Loss: 0.220083

Test set: Average loss: 0.1098, Accuracy: 9677/10000 (97%)

Train Epoch: 10 [0/60000 (0%)]  Loss: 0.194172
Train Epoch: 10 [32000/60000 (53%)]     Loss: 0.196318

Test set: Average loss: 0.0498, Accuracy: 9831/10000 (98%)

Train Epoch: 11 [0/60000 (0%)]  Loss: 0.095472
Train Epoch: 11 [32000/60000 (53%)]     Loss: 0.325434

Test set: Average loss: 0.0459, Accuracy: 9859/10000 (99%)

Train Epoch: 12 [0/60000 (0%)]  Loss: 0.125588
Train Epoch: 12 [32000/60000 (53%)]     Loss: 0.038847

Test set: Average loss: 0.0403, Accuracy: 9877/10000 (99%)

Train Epoch: 13 [0/60000 (0%)]  Loss: 0.153203
Train Epoch: 13 [32000/60000 (53%)]     Loss: 0.055198

Test set: Average loss: 0.0503, Accuracy: 9839/10000 (98%)

Train Epoch: 14 [0/60000 (0%)]  Loss: 0.209414
Train Epoch: 14 [32000/60000 (53%)]     Loss: 0.070203

Test set: Average loss: 0.0462, Accuracy: 9854/10000 (99%)

Train Epoch: 15 [0/60000 (0%)]  Loss: 0.109152
Train Epoch: 15 [32000/60000 (53%)]     Loss: 0.105076

Test set: Average loss: 0.0431, Accuracy: 9863/10000 (99%)

Train Epoch: 16 [0/60000 (0%)]  Loss: 0.036552
Train Epoch: 16 [32000/60000 (53%)]     Loss: 0.062231

Test set: Average loss: 0.0416, Accuracy: 9878/10000 (99%)

Train Epoch: 17 [0/60000 (0%)]  Loss: 0.132890
Train Epoch: 17 [32000/60000 (53%)]     Loss: 0.037140

Test set: Average loss: 0.0388, Accuracy: 9876/10000 (99%)

Train Epoch: 18 [0/60000 (0%)]  Loss: 0.107907
Train Epoch: 18 [32000/60000 (53%)]     Loss: 0.071968

Test set: Average loss: 0.0520, Accuracy: 9842/10000 (98%)

Train Epoch: 19 [0/60000 (0%)]  Loss: 0.065074
Train Epoch: 19 [32000/60000 (53%)]     Loss: 0.144325

Test set: Average loss: 0.0396, Accuracy: 9879/10000 (99%)

Train Epoch: 20 [0/60000 (0%)]  Loss: 0.040710
Train Epoch: 20 [32000/60000 (53%)]     Loss: 0.013562

Test set: Average loss: 0.0358, Accuracy: 9898/10000 (99%)