main.py

# This code is come from MorvanZhou, thank him! And I change some code.
# https://github.com/MorvanZhou/PyTorch-Tutorial/blob/master/tutorial-contents/401_CNN.py

import torch
import torch.nn as nn
from torch.autograd import Variable
import torch.utils.data as Data
import torchvision
import config as cfg
from lib.conv_offset2D import ConvOffset2D

torch.manual_seed(1)    # reproducible

# Hyper Parameters
EPOCH = 5                      # train the training data n times, to save time, we just train 1 epoch
BATCH_SIZE = 50
LR = 0.001              # learning rate
DOWNLOAD_MNIST = True   # set to False if you have downloaded


# Mnist digits dataset
train_data = torchvision.datasets.MNIST(
    root='./mnist/',
    train=True,                                     # this is training data
    transform=torchvision.transforms.ToTensor(),    # Converts a PIL.Image or numpy.ndarray to
                                                    # torch.FloatTensor of shape (C x H x W) and normalize in the range [0.0, 1.0]
    download=DOWNLOAD_MNIST,                        # download it if you don't have it
)

# Data Loader for easy mini-batch return in training, the image batch shape will be (50, 1, 28, 28)
train_loader = Data.DataLoader(dataset=train_data, batch_size=BATCH_SIZE, shuffle=True)

# convert test data into Variable, pick 2000 samples to speed up testing
test_data = torchvision.datasets.MNIST(root='./mnist/', train=False, transform=torchvision.transforms.ToTensor())

test_loader = Data.DataLoader(dataset=test_data, batch_size=2000, shuffle=False)
test_x, test_y = test_loader.__iter__().next()
test_x = Variable(test_x, volatile=True)


class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = nn.Sequential(         # input shape (1, 28, 28)
            nn.Conv2d(
                in_channels=1,              # input height
                out_channels=16,            # n_filters
                kernel_size=5,              # filter size
                stride=1,                   # filter movement/step
                padding=2,                  # if want same width and length of this image after con2d, padding=(kernel_size-1)/2 if stride=1
            ),                              # output shape (16, 28, 28)
            nn.ReLU(),                      # activation
            nn.MaxPool2d(kernel_size=2),    # choose max value in 2x2 area, output shape (16, 14, 14)
        )
        self.conv2 = nn.Sequential(         # input shape (1, 28, 28)
            nn.Conv2d(16, 32, 5, 1, 2),     # output shape (32, 14, 14)
            nn.ReLU(),                      # activation
            nn.MaxPool2d(2),                # output shape (32, 7, 7)
        )
        self.out = nn.Linear(32 * 7 * 7, 10)   # fully connected layer, output 10 classes

    def forward(self, x):
        x = self.conv1(x)
        x = self.conv2(x)
        x = x.view(x.size(0), -1)           # flatten the output of conv2 to (batch_size, 32 * 7 * 7)
        output = self.out(x)
        return output


class Net_Deform(nn.Module):
    def __init__(self):
        super(Net_Deform, self).__init__()
        self.conv1 = nn.Sequential(         # input shape (1, 28, 28)
            ConvOffset2D(filters=1),
            nn.Conv2d(1, 16, 5, 1, 2),
            nn.ReLU(),
            nn.MaxPool2d(kernel_size=2),
        )
        self.conv2 = nn.Sequential(
            ConvOffset2D(filters=16),
            nn.Conv2d(16, 32, 5, 1, 2),
            nn.ReLU(),
            nn.MaxPool2d(2),
        )
        self.out = nn.Linear(32 * 7 * 7, 10)

    def forward(self, x):
        x = self.conv1(x)
        x = self.conv2(x)
        x = x.view(x.size(0), -1)           # flatten the output of conv2 to (batch_size, 32 * 7 * 7)
        # print(x.size())

        output = self.out(x)
        return output

net = Net_Deform()
# net = Net()

if torch.cuda.is_available():
    net.cuda(cfg.cuda_num)
    test_x = test_x.cuda(cfg.cuda_num)
    test_y = test_y.cuda(cfg.cuda_num)

optimizer = torch.optim.Adam(net.parameters(), lr=LR)
loss_func = nn.CrossEntropyLoss()


# training and testing
for epoch in range(EPOCH):
    for step, (x, y) in enumerate(train_loader):
        b_x = Variable(x)
        b_y = Variable(y)

        if torch.cuda.is_available():
            b_x = b_x.cuda(cfg.cuda_num)
            b_y = b_y.cuda(cfg.cuda_num)

        net.train()
        output = net(b_x)
        loss = loss_func(output, b_y)
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        # print('(Epoch:%d|Step:%d )' % (epoch, step), '| train loss: %.4f' % loss.data[0],)

        if step % 50 == 0:
            net.eval()
            test_output = net(test_x)
            pred_y = torch.max(test_output, 1)[1].data.squeeze()
            accuracy = sum(pred_y == test_y) / float(test_y.size(0))
            print('(Epoch:%d|Step:%d )' % (epoch, step), '| train loss: %.4f' % loss.data[0], '| test accuracy: %.4f' % accuracy)

# print 10 predictions from test data
test_output = net(test_x[:10])
pred_y = torch.max(test_output, 1)[1].cpu().data.numpy().squeeze()
print(pred_y, 'prediction number')
print(test_y[:10].cpu().numpy(), 'real number')