\n Generative Adversarial Networks (GAN)\uc740 Ian Goodfellow\uac00 2014\ub144 \uc81c\uc548\ud55c \uae30\ubcf8\uc801\uc778 \ub525\ub7ec\ub2dd \ubaa8\ub378 \uc911 \ud558\ub098\uc785\ub2c8\ub2e4. GAN\uc740 \ub450 \uac1c\uc758 \uc2e0\uacbd\ub9dd\uc73c\ub85c \uad6c\uc131\ub429\ub2c8\ub2e4: \n GAN\uc758 \uad6c\uc870\ub294 \ub2e4\uc74c\uacfc \uac19\uc774 \uc791\ub3d9\ud569\ub2c8\ub2e4:\n <\/p>\n \n GAN\uc758 \ud6c8\ub828 \uacfc\uc815\uc740 \ub2e4\uc74c\uacfc \uac19\uc740 \ubc18\ubcf5\uc801\uc778 \ub2e8\uacc4\ub85c \uc774\ub8e8\uc5b4\uc838 \uc788\uc2b5\ub2c8\ub2e4:\n <\/p>\n \n \uc774\uc81c GAN\uc744 \ud30c\uc774\ud1a0\uce58\ub85c \uad6c\ud604\ud574\ubcf4\uaca0\uc2b5\ub2c8\ub2e4. \uc6b0\ub9ac\uc758 \ubaa9\ud45c\ub294 MNIST \ub370\uc774\ud130\uc14b\uc744 \uc0ac\uc6a9\ud558\uc5ec \uc22b\uc790 \uc774\ubbf8\uc9c0\ub97c \uc0dd\uc131\ud558\ub294 \uac83\uc785\ub2c8\ub2e4.\n <\/p>\n \n MNIST \ub370\uc774\ud130\uc14b\uc740 \ud30c\uc774\ud1a0\uce58\uc758 torchvision \ub77c\uc774\ube0c\ub7ec\ub9ac\ub97c \ud1b5\ud574 \uc27d\uac8c \uac00\uc838\uc62c \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc544\ub798 \ucf54\ub4dc\ub294 \ub370\uc774\ud130\ub97c \ub85c\ub4dc\ud558\uace0 \uc804\ucc98\ub9ac\ud558\ub294 \uacfc\uc815\uc785\ub2c8\ub2e4.\n <\/p>\n \n \uc774\uc81c GAN\uc758 \ub450 \ub124\ud2b8\uc6cc\ud06c\ub97c \uc815\uc758\ud569\ub2c8\ub2e4. \uac04\ub2e8\ud55c \uc2e0\uacbd\ub9dd \uad6c\uc870\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc0dd\uc131\uc790\uc640 \uad6c\ubcc4\uc790\ub97c \ub9cc\ub4ed\ub2c8\ub2e4.\n <\/p>\n \n \uc190\uc2e4 \ud568\uc218\ub294 \uc0dd\uc131\uc790\uc640 \uad6c\ubcc4\uc790 \ubaa8\ub450\uc5d0 \ub300\ud574 \uc774\uc9c4 \uad50\ucc28 \uc5d4\ud2b8\ub85c\ud53c \uc190\uc2e4\uc744 \uc0ac\uc6a9\ud569\ub2c8\ub2e4. \ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998\uc73c\ub85c\ub294 Adam\uc744 \ucc44\ud0dd\ud569\ub2c8\ub2e4.\n <\/p>\n \n \ud6c8\ub828\uc740 \ub2e4\uc74c\uacfc \uac19\uc740 \uacfc\uc815\uc73c\ub85c \uc9c4\ud589\ub429\ub2c8\ub2e4:\n <\/p>\n \n \ud6c8\ub828\uc774 \uc644\ub8cc\ub41c \ud6c4 \uc0dd\uc131\ub41c \uc774\ubbf8\uc9c0\ub97c \uc2dc\uac01\ud654\ud569\ub2c8\ub2e4.\n <\/p>\n \n GAN\uc740 \uc774\ubbf8\uc9c0 \uc0dd\uc131 \uc678\uc5d0\ub3c4 \ub2e4\uc591\ud55c \ubd84\uc57c\uc5d0 \ud65c\uc6a9\ub420 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc608\ub97c \ub4e4\uc5b4, \uc2a4\ud0c0\uc77c \uc804\uc774, \uc774\ubbf8\uc9c0 \ubcf5\uc6d0, \ube44\ub514\uc624 \uc0dd\uc131 \ub4f1\uc5d0\uc11c \uc0ac\uc6a9\ub418\uace0 \uc788\uc73c\uba70, \uc778\uacf5\uc9c0\ub2a5 \ubd84\uc57c\uc5d0\uc11c \ud070 \uc8fc\ubaa9\uc744 \ubc1b\uace0 \uc788\uc2b5\ub2c8\ub2e4. \n \uc774\ubc88 \uac15\uc88c\uc5d0\uc11c\ub294 \ud30c\uc774\ud1a0\uce58\ub97c \uc0ac\uc6a9\ud558\uc5ec GAN\uc744 \uad6c\ud604\ud558\ub294 \uae30\ubcf8\uc801\uc778 \ubc29\ubc95\uc5d0 \ub300\ud574 \ubc30\uc6b0\uace0 \uc2e4\uc81c \ucf54\ub4dc\ub97c \ud1b5\ud574 GAN\uc758 \uc791\ub3d9 \uc6d0\ub9ac\ub97c \uc774\ud574\ud588\uc2b5\ub2c8\ub2e4. GAN\uc740 \uc55e\uc73c\ub85c\ub3c4 \ub354\uc6b1 \ubc1c\uc804\ud560 \uae30\uc220\uc774\uba70, \ub2e4\uc591\ud55c \uc751\uc6a9 \ud504\ub85c\uadf8\ub7a8\uc5d0\uc11c \ud070 \uac00\ub2a5\uc131\uc744 \uc9c0\ub2c8\uace0 \uc788\uc2b5\ub2c8\ub2e4. <\/body><\/p>\n","protected":false},"excerpt":{"rendered":"
\n \uc0dd\uc131\uc790(Generator)<\/strong>\uc640 \uad6c\ubcc4\uc790(Discriminator)<\/strong>. \uc0dd\uc131\uc790\ub294 \uac00\uc9dc \ub370\uc774\ud130\ub97c \uc0dd\uc131\ud558\ub824 \ud558\uace0, \uad6c\ubcc4\uc790\ub294 \uc774 \ub370\uc774\ud130\uac00 \uc9c4\uc9dc\uc778\uc9c0 \uac00\uc9dc\uc778\uc9c0 \ud310\ub2e8\ud558\ub824 \ud569\ub2c8\ub2e4.
\n \uc774 \ub450 \ub124\ud2b8\uc6cc\ud06c\ub294 \uc11c\ub85c \uacbd\uc7c1\ud558\uba70 \ud6c8\ub828\ub418\uae30 \ub54c\ubb38\uc5d0 “Adversarial”\uc774\ub77c\ub294 \uc6a9\uc5b4\uac00 \ubd99\uc2b5\ub2c8\ub2e4. \uc774 \uacfc\uc815\uc740 \ube44\uc9c0\ub3c4 \ud559\uc2b5 \ubc29\uc2dd\uc73c\ub85c, \uc0dd\uc131\uc790\ub294 \uc810\uc810 \ub354 \uc9c4\uc9dc \ub370\uc774\ud130\uc640 \ube44\uc2b7\ud55c \ub370\uc774\ud130\ub97c \uc0dd\uc131\ud558\uac8c \ub429\ub2c8\ub2e4.\n <\/p>\n2. GAN\uc758 \uad6c\uc870<\/h2>\n
\n
3. GAN\uc758 \ub3d9\uc791 \uc6d0\ub9ac<\/h2>\n
\n
4. GAN\uc744 \ud30c\uc774\ud1a0\uce58\ub85c \uad6c\ud604\ud558\uae30<\/h2>\n
4.1 \ud544\uc694\ud55c \ub77c\uc774\ube0c\ub7ec\ub9ac \uc124\uce58<\/h3>\n
\n
pip install torch torchvision matplotlib<\/code>\n <\/pre>\n4.2 \ub370\uc774\ud130\uc14b \uc900\ube44\ud558\uae30<\/h3>\n
\n
\nimport torch\nfrom torchvision import datasets, transforms\nfrom torch.utils.data import DataLoader\n\n# \ub370\uc774\ud130 \ubcc0\ud658 \uc124\uc815\ntransform = transforms.Compose([\n transforms.ToTensor(),\n transforms.Normalize((0.5,), (0.5,))\n])\n\n# MNIST \ub370\uc774\ud130\uc14b \ub85c\ub4dc\ntrain_dataset = datasets.MNIST(root='.\/data', train=True, download=True, transform=transform)\ntrain_loader = DataLoader(train_dataset, batch_size=64, shuffle=True)\n <\/code>\n <\/pre>\n4.3 \uc0dd\uc131\uc790\uc640 \uad6c\ubcc4\uc790 \ub124\ud2b8\uc6cc\ud06c \uc815\uc758<\/h3>\n
\n
\nimport torch.nn as nn\n\n# \uc0dd\uc131\uc790 \ub124\ud2b8\uc6cc\ud06c \uc815\uc758\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, 28 * 28),\n nn.Tanh()\n )\n\n def forward(self, z):\n return self.model(z).view(-1, 1, 28, 28)\n\n# \uad6c\ubcc4\uc790 \ub124\ud2b8\uc6cc\ud06c \uc815\uc758\nclass Discriminator(nn.Module):\n def __init__(self):\n super(Discriminator, self).__init__()\n self.model = nn.Sequential(\n nn.Flatten(),\n nn.Linear(28 * 28, 512),\n nn.LeakyReLU(0.2),\n nn.Linear(512, 256),\n nn.LeakyReLU(0.2),\n nn.Linear(256, 1),\n nn.Sigmoid()\n )\n\n def forward(self, img):\n return self.model(img)\n <\/code>\n <\/pre>\n4.4 \uc190\uc2e4 \ud568\uc218\uc640 \ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998 \uc124\uc815<\/h3>\n
\n
\nimport torch.optim as optim\n\n# \uc190\uc2e4 \ud568\uc218 \ubc0f \ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998 \uc815\uc758\ncriterion = nn.BCELoss()\ngenerator = Generator()\ndiscriminator = Discriminator()\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>\n4.5 \ud6c8\ub828 \uacfc\uc815 \uad6c\ud604<\/h3>\n
\n
\nimport numpy as np\n\nnum_epochs = 50\nfor epoch in range(num_epochs):\n for i, (images, _) in enumerate(train_loader):\n batch_size = images.size(0)\n \n # \ud604\uc2e4\uacfc \ube44\ud604\uc2e4 \ud0dc\uadf8 \uc124\uc815\n real_labels = torch.ones(batch_size, 1)\n fake_labels = torch.zeros(batch_size, 1)\n\n # \uad6c\ubcc4\uc790 \ud6c8\ub828\n outputs = discriminator(images)\n d_loss_real = criterion(outputs, real_labels)\n\n noise = torch.randn(batch_size, 100)\n fake_images = generator(noise)\n outputs = discriminator(fake_images.detach())\n d_loss_fake = criterion(outputs, fake_labels)\n\n d_loss = d_loss_real + d_loss_fake\n optimizer_D.zero_grad()\n d_loss.backward()\n optimizer_D.step()\n\n # \uc0dd\uc131\uc790 \ud6c8\ub828\n outputs = discriminator(fake_images)\n g_loss = criterion(outputs, real_labels)\n\n optimizer_G.zero_grad()\n g_loss.backward()\n optimizer_G.step()\n\n # \ud6c8\ub828 \uc0c1\ud0dc \ucd9c\ub825\n if (i + 1) % 100 == 0:\n print(f'Epoch [{epoch + 1}\/{num_epochs}], Step [{i + 1}\/{len(train_loader)}], d_loss: {d_loss.item():.4f}, g_loss: {g_loss.item():.4f}')\n <\/code>\n <\/pre>\n4.6 \uc0dd\uc131\ub41c \uc774\ubbf8\uc9c0 \uc2dc\uac01\ud654\ud558\uae30<\/h3>\n
\n
\nimport matplotlib.pyplot as plt\n\ndef visualize_generator(num_images):\n noise = torch.randn(num_images, 100)\n with torch.no_grad():\n generated_images = generator(noise)\n\n plt.figure(figsize=(10, 10))\n for i in range(num_images):\n plt.subplot(5, 5, i + 1)\n plt.imshow(generated_images[i][0].cpu().numpy(), cmap='gray')\n plt.axis('off')\n plt.show()\n\nvisualize_generator(25)\n <\/code>\n <\/pre>\n5. GAN\uc758 \ud65c\uc6a9<\/h2>\n
\n GAN\uc758 \ubc1c\uc804\uc740 \uc0dd\uc131\uc801 \ubaa8\ub378\uc744 \ud1b5\ud574 \ubb34\uc5c7\uc774 \uac00\ub2a5\ud55c\uc9c0\ub97c \uc0c8\ub86d\uac8c \ubcf4\uc5ec\uc8fc\uace0 \uc788\uc2b5\ub2c8\ub2e4.\n <\/p>\n6. \uacb0\ub860<\/h2>\n
\n \uc55e\uc73c\ub85c \ub354 \ub098\uc544\uac00 GAN\uc744 \uae30\ubc18\uc73c\ub85c \ud55c \ub2e4\uc591\ud55c \ubcc0\ud615 \ubaa8\ub378\ub4e4\uc744 \ud0d0\uad6c\ud574\ubcf4\uae38 \ucd94\ucc9c\ud569\ub2c8\ub2e4.\n <\/p>\n