Discriminator<\/strong>: \uc2e4\uc81c \uc0d8\ud50c(x)\uc640 \uc0dd\uc131\ub41c \uc0d8\ud50c(x’)\uc744 \uc785\ub825\ubc1b\uc544 \uc774\ub4e4\uc774 \uc2e4\uc81c\uc778\uc9c0 \uc0dd\uc131\ub41c \uac83\uc778\uc9c0 \ud310\ub2e8\ud569\ub2c8\ub2e4.<\/li>\n<\/ul>\n\n GAN\uc740 \uacb0\uad6d \uc0dd\uc131\uc790\uac00 \uc0dd\uc131\ud55c \ub370\uc774\ud130\uac00 \uc2e4\uc81c \ub370\uc774\ud130\uc640 \uad6c\ubd84\ub418\uc9c0 \uc54a\ub3c4\ub85d \ub9cc\ub4dc\ub294 \uac83\uc774 \ubaa9\ud45c\uc785\ub2c8\ub2e4.\n <\/p>\n
3. \ud30c\uc774\ud1a0\uce58\ub97c \ud65c\uc6a9\ud55c GAN \uad6c\ud604<\/h2>\n
\n \ud30c\uc774\ud1a0\uce58\ub294 \ub525\ub7ec\ub2dd \ubaa8\ub378\uc744 \uad6c\ud604\ud558\ub294 \ub370 \ub9e4\uc6b0 \uc720\uc6a9\ud55c \ud504\ub808\uc784\uc6cc\ud06c\uc785\ub2c8\ub2e4. \ub2e4\uc74c\uc740 \ud30c\uc774\ud1a0\uce58\ub97c \ud65c\uc6a9\ud558\uc5ec \uac04\ub2e8\ud55c GAN\uc744 \uad6c\ud604\ud558\ub294 \uc608\uc81c\uc785\ub2c8\ub2e4. \uc774\ubc88 \uc608\uc81c\uc5d0\uc11c\ub294 MNIST \ub370\uc774\ud130\uc14b\uc744 \uc0ac\uc6a9\ud558\uc5ec \uc190\uae00\uc528 \uc22b\uc790\ub97c \uc0dd\uc131\ud558\ub294 GAN \ubaa8\ub378\uc744 \uad6c\ucd95\ud558\uaca0\uc2b5\ub2c8\ub2e4.\n <\/p>\n
3.1 \ud658\uacbd \uc124\uc815<\/h3>\n
\n \uba3c\uc800 \ud544\uc694\ud55c \ub77c\uc774\ube0c\ub7ec\ub9ac\ub97c \uc124\uce58\ud569\ub2c8\ub2e4. \uc544\ub798\uc758 \ucf54\ub4dc\ub97c \uc0ac\uc6a9\ud558\uc5ec \ud30c\uc774\ud1a0\uce58\uc640 torchvision\uc744 \uc124\uce58\ud569\ub2c8\ub2e4.\n <\/p>\n
\n \npip install torch torchvision\n <\/code>\n <\/pre>\n3.2 \ub370\uc774\ud130\uc14b \ub85c\ub4dc<\/h3>\n
\n MNIST \ub370\uc774\ud130\uc14b\uc744 \ub2e4\uc6b4\ub85c\ub4dc\ud558\uace0 \ub85c\ub4dc\ud569\ub2c8\ub2e4. \ub2e4\uc74c \ucf54\ub4dc\ub97c \uc0ac\uc6a9\ud558\uc5ec \ub370\uc774\ud130\uc14b\uc744 \uc900\ube44\ud569\ub2c8\ub2e4.\n <\/p>\n
\n \nimport torch\nfrom torchvision import datasets, transforms\n\n# \ub370\uc774\ud130\uc14b \ubcc0\ud658\ntransform = transforms.Compose([\n transforms.ToTensor(),\n transforms.Normalize((0.5,), (0.5,))\n])\n\n# MNIST \ub370\uc774\ud130\uc14b \ub2e4\uc6b4\ub85c\ub4dc\nmnist_dataset = datasets.MNIST(root='.\/data', train=True, download=True, transform=transform)\n\n# \ub370\uc774\ud130\ub85c\ub354 \uc124\uc815\ndataloader = torch.utils.data.DataLoader(mnist_dataset, batch_size=64, shuffle=True)\n <\/code>\n <\/pre>\n3.3 \uc0dd\uc131\uc790 \ubaa8\ub378 \uc815\uc758<\/h3>\n
\n \uc0dd\uc131\uc790 \ubaa8\ub378\uc740 \ub79c\ub364 \uc7a0\uc7ac \ubca1\ud130\ub97c \uc785\ub825\ubc1b\uc544 \uc774\ubbf8\uc9c0\ub97c \uc0dd\uc131\ud558\ub294 \uc5ed\ud560\uc744 \ud569\ub2c8\ub2e4. \uc544\ub798\ub294 \uac04\ub2e8\ud55c \uc0dd\uc131\uc790 \ubaa8\ub378\uc744 \uc815\uc758\ud558\ub294 \ucf54\ub4dc\uc785\ub2c8\ub2e4.\n <\/p>\n
\n \nimport torch.nn as nn\n\nclass Generator(nn.Module):\n def __init__(self):\n super(Generator, self).__init__()\n self.model = nn.Sequential(\n nn.Linear(100, 256),\n nn.ReLU(),\n nn.Linear(256, 512),\n nn.ReLU(),\n nn.Linear(512, 1024),\n nn.ReLU(),\n nn.Linear(1024, 784), # 28x28 \uc774\ubbf8\uc9c0\ub85c \ucd9c\ub825\n nn.Tanh() # \uc785\ub825 \ubc94\uc704\ub97c [-1, 1]\ub85c \uc870\uc815\n )\n\n def forward(self, z):\n return self.model(z)\n <\/code>\n <\/pre>\n3.4 \ud310\ubcc4\uc790 \ubaa8\ub378 \uc815\uc758<\/h3>\n
\n \ud310\ubcc4\uc790 \ubaa8\ub378\uc740 \uc785\ub825 \ub370\uc774\ud130\ub97c \ud3c9\uac00\ud558\uc5ec \uc2e4\uc81c\uc778\uc9c0 \uac00\uc9dc\uc778\uc9c0 \ud310\ub2e8\ud569\ub2c8\ub2e4. \ub2e4\uc74c \ucf54\ub4dc\uc5d0\uc11c \ud310\ubcc4\uc790 \ubaa8\ub378\uc744 \uc815\uc758\ud569\ub2c8\ub2e4.\n <\/p>\n
\n \nclass Discriminator(nn.Module):\n def __init__(self):\n super(Discriminator, self).__init__()\n self.model = nn.Sequential(\n nn.Linear(784, 512), # 28x28 \uc774\ubbf8\uc9c0\ub85c\ubd80\ud130 784 \ucc28\uc6d0\n nn.LeakyReLU(0.2),\n nn.Linear(512, 256),\n nn.LeakyReLU(0.2),\n nn.Linear(256, 1), # \ucd5c\uc885 \ucd9c\ub825\uc744 1\ub85c \uc124\uc815 (\uc2e4\uc81c\/\uac00\uc9dc \ud310\ub2e8)\n nn.Sigmoid() # \ucd9c\ub825 \ubc94\uc704\ub97c [0, 1]\ub85c \uc870\uc815\n )\n\n def forward(self, x):\n return self.model(x)\n <\/code>\n <\/pre>\n3.5 \uc190\uc2e4 \ud568\uc218 \ubc0f \uc635\ud2f0\ub9c8\uc774\uc800 \uc124\uc815<\/h3>\n
\n GAN\uc758 \uc190\uc2e4 \ud568\uc218\ub85c\ub294 Binary Cross Entropy\ub97c \uc0ac\uc6a9\ud558\uba70, \uac01\uac01\uc758 \ub124\ud2b8\uc6cc\ud06c\uc5d0 \ub300\ud574 \uc635\ud2f0\ub9c8\uc774\uc800\ub97c \uc815\uc758\ud569\ub2c8\ub2e4. \ub2e4\uc74c \ucf54\ub4dc\ub97c \uc0ac\uc6a9\ud569\ub2c8\ub2e4.\n <\/p>\n
\n \nimport torch.optim as optim\n\n# \ubaa8\ub378 \uc778\uc2a4\ud134\uc2a4 \uc0dd\uc131\ngenerator = Generator()\ndiscriminator = Discriminator()\n\n# \uc190\uc2e4 \ud568\uc218 \ubc0f \uc635\ud2f0\ub9c8\uc774\uc800 \uc124\uc815\ncriterion = nn.BCELoss()\noptimizer_G = optim.Adam(generator.parameters(), lr=0.0002, betas=(0.5, 0.999))\noptimizer_D = optim.Adam(discriminator.parameters(), lr=0.0002, betas=(0.5, 0.999))\n <\/code>\n <\/pre>\n3.6 GAN \ud559\uc2b5 \ub8e8\ud504<\/h3>\n
\n \ubaa8\ub378\uc744 \ud559\uc2b5\uc2dc\ud0a4\uae30 \uc704\ud55c \ub8e8\ud504\ub97c \uc791\uc131\ud569\ub2c8\ub2e4. \uac01 \ubc18\ubcf5\uc5d0\uc11c \uc0dd\uc131\uc790\ub294 \uac00\uc9dc \uc0d8\ud50c\uc744 \uc0dd\uc131\ud558\uace0, \ud310\ubcc4\uc790\ub294 \uc774\ub97c \ud3c9\uac00\ud558\uc5ec \uc190\uc2e4\uc744 \uacc4\uc0b0\ud569\ub2c8\ub2e4.\n <\/p>\n
\n \nnum_epochs = 200\n\nfor epoch in range(num_epochs):\n for i, (images, _) in enumerate(dataloader):\n # \ubc30\uce58 \ud06c\uae30 \uc124\uc815\n batch_size = images.size(0)\n \n # \ub77c\ubca8 \uc0dd\uc131\n real_labels = torch.ones(batch_size, 1)\n fake_labels = torch.zeros(batch_size, 1)\n \n # \ud310\ubcc4\uc790 \ud559\uc2b5\n optimizer_D.zero_grad()\n \n # \uc2e4\uc81c \uc774\ubbf8\uc9c0\uc5d0 \ub300\ud55c \uc190\uc2e4\n outputs = discriminator(images.view(batch_size, -1))\n d_loss_real = criterion(outputs, real_labels)\n \n # \uac00\uc9dc \uc774\ubbf8\uc9c0 \uc0dd\uc131\n z = torch.randn(batch_size, 100)\n fake_images = generator(z)\n \n # \uac00\uc9dc \uc774\ubbf8\uc9c0\uc5d0 \ub300\ud55c \uc190\uc2e4\n outputs = discriminator(fake_images.detach())\n d_loss_fake = criterion(outputs, fake_labels)\n \n # \ucd1d \ud310\ubcc4\uc790 \uc190\uc2e4\n d_loss = d_loss_real + d_loss_fake\n d_loss.backward()\n optimizer_D.step()\n \n # \uc0dd\uc131\uc790 \ud559\uc2b5\n optimizer_G.zero_grad()\n outputs = discriminator(fake_images)\n g_loss = criterion(outputs, real_labels)\n g_loss.backward()\n optimizer_G.step()\n \n # \uc5d0\ud3ec\ud06c \ud6c4 \uc190\uc2e4 \ucd9c\ub825\n if (epoch + 1) % 10 == 0:\n print(f'Epoch [{epoch + 1}\/{num_epochs}], d_loss: {d_loss.item():.4f}, g_loss: {g_loss.item():.4f}')\n <\/code>\n <\/pre>\n3.7 \uacb0\uacfc \uc2dc\uac01\ud654<\/h3>\n
\n \uc0dd\uc131\ub41c \uc774\ubbf8\uc9c0\ub97c \uc2dc\uac01\ud654\ud558\uae30 \uc704\ud574 Matplotlib\uc744 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \ub2e4\uc74c \ucf54\ub4dc\ub97c \ud1b5\ud574 \uc774\ubbf8\uc9c0\ub97c \uc2dc\uac01\ud654\ud569\ub2c8\ub2e4.\n <\/p>\n
\n \nimport matplotlib.pyplot as plt\n\n# \uc0dd\uc131\ud55c \uc774\ubbf8\uc9c0\ub97c \uc2dc\uac01\ud654\ndef visualize_images(generator, num_images=64):\n z = torch.randn(num_images, 100)\n fake_images = generator(z).view(-1, 1, 28, 28).detach()\n \n grid = torchvision.utils.make_grid(fake_images, nrow=8, normalize=True)\n plt.imshow(grid.permute(1, 2, 0).numpy())\n plt.axis('off')\n plt.show()\n\n# \uc608\uc2dc \uc774\ubbf8\uc9c0 \uc2dc\uac01\ud654\nvisualize_images(generator, 64)\n <\/code>\n <\/pre>\n4. GAN\uc758 \uc801\uc6a9 \ubd84\uc57c<\/h2>\n
\n GAN\uc740 \uc5ec\ub7ec \ubd84\uc57c\uc5d0\uc11c \uadf8 \uac00\ub2a5\uc131\uc744 \ubcf4\uc5ec\uc8fc\uace0 \uc788\uc2b5\ub2c8\ub2e4. \ub2e4\uc74c\uc740 GAN\uc758 \uc8fc\uc694 \uc801\uc6a9 \ubd84\uc57c\uc785\ub2c8\ub2e4.\n <\/p>\n
4.1 \uc774\ubbf8\uc9c0 \uc0dd\uc131<\/h3>\n
\n GAN\uc740 \uace0\ud488\uc9c8 \uc774\ubbf8\uc9c0\ub97c \uc0dd\uc131\ud558\ub294 \ub370 \ud65c\uc6a9\ub429\ub2c8\ub2e4. \uc608\ub97c \ub4e4\uc5b4, DCGAN(Deep Convolutional GAN)\uc740 \uc2e4\uc81c\ucc98\ub7fc \ubcf4\uc774\ub294 \uc774\ubbf8\uc9c0\ub97c \uc0dd\uc131\ud558\ub294 \ub370 \ub110\ub9ac \uc0ac\uc6a9\ub429\ub2c8\ub2e4.\n <\/p>\n
4.2 \uc2a4\ud0c0\uc77c \ubcc0\ud658<\/h3>\n
\n GAN\uc740 \uc774\ubbf8\uc9c0 \uc2a4\ud0c0\uc77c\uc744 \ubcc0\ud658\ud558\ub294 \ub370\uc5d0\ub3c4 \uc0ac\uc6a9\ub429\ub2c8\ub2e4. CycleGAN\uacfc \uac19\uc740 \ubaa8\ub378\uc740 \ud2b9\uc815 \uc2a4\ud0c0\uc77c\uc758 \uc774\ubbf8\uc9c0\ub97c \ub2e4\ub978 \uc2a4\ud0c0\uc77c\ub85c \ubcc0\ud658\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc608\ub97c \ub4e4\uc5b4, \uc5ec\ub984 \ud48d\uacbd\uc744 \uaca8\uc6b8 \ud48d\uacbd\uc73c\ub85c \ubcc0\ud658\ud558\ub294 \uac83\uc774 \uac00\ub2a5\ud569\ub2c8\ub2e4.\n <\/p>\n
4.3 \uc774\ubbf8\uc9c0 \ubcf4\uc644 \ubc0f \uc288\ud37c \ud574\uc0c1\ub3c4<\/h3>\n
\n GAN\uc740 \uc774\ubbf8\uc9c0 \ub0b4 \uacb0\ud568\uc744 \ubcf4\uc644\ud558\uac70\ub098 \uc800\ud574\uc0c1\ub3c4\ub97c \uace0\ud574\uc0c1\ub3c4\ub85c \ubcc0\ud658\ud558\ub294 \ub370 \uc0ac\uc6a9\ub420 \uc218 \uc788\uc2b5\ub2c8\ub2e4. SRGAN(Super Resolution GAN)\uc740 \uc800\ud574\uc0c1\ub3c4 \uc774\ubbf8\uc9c0\ub97c \uace0\ud574\uc0c1\ub3c4 \uc774\ubbf8\uc9c0\ub85c \ubcc0\ud658\ud569\ub2c8\ub2e4.\n <\/p>\n
4.4 \ube44\ub514\uc624 \uc0dd\uc131<\/h3>\n
\n GAN\uc740 \uc774\ubbf8\uc9c0\ubfd0\ub9cc \uc544\ub2c8\ub77c \ube44\ub514\uc624 \uc0dd\uc131\uc5d0\ub3c4 \ud65c\uc6a9\ub429\ub2c8\ub2e4. MovGAN\uacfc \uac19\uc740 \ubaa8\ub378\uc740 \uc5f0\uc18d\uc801\uc778 \ud504\ub808\uc784\uc744 \uc0dd\uc131\ud558\uc5ec \ub9ac\uc5bc\ud55c \ube44\ub514\uc624 \uc2dc\ud000\uc2a4\ub97c \ub9cc\ub4ed\ub2c8\ub2e4.\n <\/p>\n
4.5 \uc790\uc5f0\uc5b4 \ucc98\ub9ac<\/h3>\n
\n GAN\uc740 \ud14d\uc2a4\ud2b8 \uc0dd\uc131\uc744 \ud3ec\ud568\ud55c \uc790\uc5f0\uc5b4 \ucc98\ub9ac(NLP)\uc5d0\uc11c\ub3c4 \uc0ac\uc6a9\ub429\ub2c8\ub2e4. TextGAN\uacfc \uac19\uc740 \ubaa8\ub378\uc740 \uc8fc\uc5b4\uc9c4 \ucee8\ud14d\uc2a4\ud2b8\uc5d0 \uae30\ubc18\ud558\uc5ec \ud14d\uc2a4\ud2b8\ub97c \uc0dd\uc131\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.\n <\/p>\n
4.6 \ub370\uc774\ud130 \uc99d\uac15<\/h3>\n
\n GAN\uc740 \ub370\uc774\ud130\uc14b\uc744 \ud655\uc7a5\ud558\ub294 \ub370 \uc0ac\uc6a9\ub420 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \ud2b9\ud788, \ud2b9\uc815 \ud074\ub798\uc2a4\uc758 \ub370\uc774\ud130\uac00 \ubd80\uc871\ud560 \ub54c \uc0dd\uc131\ub41c \uc774\ubbf8\uc9c0\ub97c \uc0ac\uc6a9\ud558\uc5ec \ub370\uc774\ud130\ub97c \ubcf4\uac15\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.\n <\/p>\n
4.7 \uc758\ub8cc \uc601\uc0c1<\/h3>\n
\n GAN\uc740 \uc758\ub8cc \ubd84\uc57c\uc5d0\uc11c\ub3c4 \ud65c\uc6a9\ub429\ub2c8\ub2e4. \uc758\ub8cc \uc601\uc0c1\uc744 \uc0dd\uc131\ud558\uace0 \uc804\ucc98\ub9ac\ud558\uc5ec \uc9c4\ub2e8 \ubcf4\uc870 \ub3c4\uad6c\ub85c \uc0ac\uc6a9\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc608\ub97c \ub4e4\uc5b4, CT \uc2a4\uce94\uc774\ub098 MRI \uc774\ubbf8\uc9c0\ub97c \uc0dd\uc131\ud558\ub294 \ub370 \uc0ac\uc6a9\ub429\ub2c8\ub2e4.\n <\/p>\n
\uacb0\ub860<\/h2>\n
\n GAN\uc740 \uc0dd\uc131 \ubaa8\ub378\ub9c1 \ubd84\uc57c\uc5d0\uc11c \ud601\uc2e0\uc801\uc778 \ubc1c\uc804\uc744 \uc774\ub8e8\uc5b4\ub0b8 \ub525\ub7ec\ub2dd \ubaa8\ub378\uc785\ub2c8\ub2e4. \ud30c\uc774\ud1a0\uce58\ub97c \ud65c\uc6a9\ud55c \uad6c\ud604\uc744 \ud1b5\ud574 GAN\uc758 \uc791\ub3d9 \uc6d0\ub9ac\uc640 \uad6c\uc870\ub97c \uc774\ud574\ud560 \uc218 \uc788\uc5c8\uc73c\uba70, \ub2e4\uc591\ud55c \uc801\uc6a9 \ubd84\uc57c\ub97c \uc0b4\ud3b4\ubcf4\uc558\uc2b5\ub2c8\ub2e4. GAN\uc758 \uac00\ub2a5\uc131\uc740 \ubb34\uad81\ubb34\uc9c4\ud558\uba70, \uc55e\uc73c\ub85c\ub3c4 \uacc4\uc18d\ud574\uc11c \ubc1c\uc804\ud560 \uac83\uc73c\ub85c \uae30\ub300\ub429\ub2c8\ub2e4. \uc774\ub7ec\ud55c \uae30\uc220\ub4e4\uc774 \uc138\uc0c1\uc5d0 \uae0d\uc815\uc801\uc778 \uc601\ud5a5\uc744 \ubbf8\uce58\uae38 \ubc14\ub77c\uba70, GAN\uc744 \ud65c\uc6a9\ud55c \ud504\ub85c\uc81d\ud2b8\uc5d0 \ub3c4\uc804\ud574\ubcf4\uc2dc\uae38 \ucd94\ucc9c\ub4dc\ub9bd\ub2c8\ub2e4.\n <\/p>\n\n