{"id":29824,"date":"2024-10-28T03:00:23","date_gmt":"2024-10-28T03:00:23","guid":{"rendered":"http:\/\/atmokpo.com\/w\/?p=29824"},"modified":"2024-11-26T06:51:13","modified_gmt":"2024-11-26T06:51:13","slug":"%ed%8c%8c%ec%9d%b4%ed%86%a0%ec%b9%98%eb%a5%bc-%ed%99%9c%ec%9a%a9%ed%95%9c-gan-%eb%94%a5%eb%9f%ac%eb%8b%9d-%ea%b3%a0%ec%95%bd%ed%95%9c-%eb%b2%94%eb%b2%95%ec%9e%90%eb%a5%bc-%ec%9c%84%ed%95%9c-%eb%ac%b8","status":"publish","type":"post","link":"https:\/\/atmokpo.com\/w\/29824\/","title":{"rendered":"\ud30c\uc774\ud1a0\uce58\ub97c \ud65c\uc6a9\ud55c GAN \ub525\ub7ec\ub2dd, \uace0\uc57d\ud55c \ubc94\ubc95\uc790\ub97c \uc704\ud55c \ubb38\ud559 \ud074\ub7fd"},"content":{"rendered":"

<\/p>\n

Generative Adversarial Networks (GANs)\ub294 \ub525\ub7ec\ub2dd\uc758 \uac00\uc7a5 \ud601\uc2e0\uc801\uc778 \ubc1c\uc804 \uc911 \ud558\ub098\ub85c \uac04\uc8fc\ub429\ub2c8\ub2e4. GAN\uc740 \ub450 \uac1c\uc758 \uc2e0\uacbd\ub9dd, \uc989 \uc0dd\uc131\uc790(Generator)\uc640 \ud310\ubcc4\uc790(Discriminator)\ub85c \uad6c\uc131\ub429\ub2c8\ub2e4. \uc0dd\uc131\uc790\ub294 \ub370\uc774\ud130\ub97c \uc0dd\uc131\ud558\uace0, \ud310\ubcc4\uc790\ub294 \ud574\ub2f9 \ub370\uc774\ud130\uac00 \uc9c4\uc9dc\uc778\uc9c0 \uac00\uc9dc\uc778\uc9c0 \ud310\ub2e8\ud569\ub2c8\ub2e4. \uc774\ub7ec\ud55c \uacbd\uc7c1 \uad6c\uc870\ub294 \uc11c\ub85c\uc758 \uc131\ub2a5\uc744 \ud5a5\uc0c1\uc2dc\ud0a4\ub294 \uc5ed\ud560\uc744 \ud569\ub2c8\ub2e4. \ubcf8 \uac15\uc88c\uc5d0\uc11c\ub294 \ud30c\uc774\ud1a0\uce58\ub97c \uc0ac\uc6a9\ud558\uc5ec GAN\uc744 \uad6c\ucd95\ud558\uace0, \uc774\ub97c \ud1b5\ud574 \ud765\ubbf8\ub85c\uc6b4 \ubc29\uc2dd\uc73c\ub85c ‘\uace0\uc57d\ud55c \ubc94\ubc95\uc790\ub97c \uc704\ud55c \ubb38\ud559 \ud074\ub7fd’\uc744 \uc8fc\uc81c\ub85c \ud55c \ub370\uc774\ud130 \uc0dd\uc131\uc744 \uc218\ud589\ud560 \uac83\uc785\ub2c8\ub2e4.<\/p>\n

1. GAN\uc758 \uae30\ubcf8 \uad6c\uc870\uc640 \uc6d0\ub9ac<\/h2>\n

GAN\uc758 \uc791\ub3d9 \ubc29\uc2dd\uc740 \ub2e4\uc74c\uacfc \uac19\uc2b5\ub2c8\ub2e4:<\/p>\n

\uc0dd\uc131\uc790(Generator)<\/strong>: \ubb34\uc791\uc704 \ub178\uc774\uc988(z)\ub97c \uc785\ub825\uc73c\ub85c \ubc1b\uc544 \ud604\uc2e4\uc801\uc778 \ub370\uc774\ud130\ub97c \uc0dd\uc131\ud569\ub2c8\ub2e4.<\/li>\n
\ud310\ubcc4\uc790(Discriminator)<\/strong>: \uc785\ub825\ub41c \ub370\uc774\ud130\uac00 \uc2e4\uc81c \ub370\uc774\ud130\uc778\uc9c0 \uc0dd\uc131\uc790\uc5d0 \uc758\ud574 \ub9cc\ub4e4\uc5b4\uc9c4 \ub370\uc774\ud130\uc778\uc9c0 \ud310\ub2e8\ud569\ub2c8\ub2e4.<\/li>\n

\uc0dd\uc131\uc790\ub294 \ud310\ubcc4\uc790\ub97c \uc18d\uc774\ub824 \ud558\uba70, \ud310\ubcc4\uc790\ub294 \uc0dd\uc131\uc790\ub97c \uad6c\ubcc4\ud558\ub824 \ud569\ub2c8\ub2e4. \uc774\ub7ec\ud55c \uacbd\uc7c1\uc774 \uc9c0\uc18d\ub418\uba74\uc11c \ub450 \ub124\ud2b8\uc6cc\ud06c\ub294 \uc810\uc810 \ub354 \ubc1c\uc804\ud558\uac8c \ub429\ub2c8\ub2e4.<\/li>\n<\/ul>\n
2. \ud544\uc694\ud55c \ub77c\uc774\ube0c\ub7ec\ub9ac\uc640 \ub370\uc774\ud130\uc14b \uc900\ube44<\/h2>\n
\ud30c\uc774\ud1a0\uce58\uc640 \uae30\ud0c0 \ud544\uc694\ud55c \ub77c\uc774\ube0c\ub7ec\ub9ac\ub97c \uc124\uce58\ud569\ub2c8\ub2e4. \uadf8\ub9ac\uace0 \ub370\uc774\ud130\ub97c \uc900\ube44\ud558\ub294 \uacfc\uc815\uc5d0\uc11c \uc0ac\uc6a9\ud560 \ub370\uc774\ud130\uc14b\uc744 \uc120\ud0dd\ud574\uc57c \ud569\ub2c8\ub2e4. \uc608\uc81c\uc5d0\uc11c\ub294 MNIST \ub370\uc774\ud130\uc14b\uc744 \uc0ac\uc6a9\ud558\uc5ec \uc22b\uc790 \uc774\ubbf8\uc9c0\ub97c \uc0dd\uc131\ud574\ubcf4\uaca0\uc2b5\ub2c8\ub2e4. MNIST \ub370\uc774\ud130\uc14b\uc740 \uc190\uc73c\ub85c \uc4f4 \uc22b\uc790 \uc774\ubbf8\uc9c0\ub85c \uad6c\uc131\ub418\uc5b4 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
2.1 \ud658\uacbd \uc124\uc815<\/h3>\n
pip install torch torchvision<\/code><\/pre>\n2.2 \ub370\uc774\ud130\uc14b \ub85c\ub4dc<\/h3>\nimport torch\nfrom torchvision import datasets, transforms\nfrom torch.utils.data import DataLoader\n\n# MNIST \ub370\uc774\ud130\uc14b \ub85c\ub4dc\ntransform = transforms.Compose([\n transforms.ToTensor(),\n transforms.Normalize((0.5,), (0.5,))\n])\n\nmnist_dataset = datasets.MNIST(root='.\/data', train=True, download=True, transform=transform)\ndata_loader = DataLoader(dataset=mnist_dataset, batch_size=64, shuffle=True)<\/code><\/pre>\n3. GAN \ubaa8\ub378 \uad6c\ucd95<\/h2>\nGenerative Adversarial Network\ub97c \uad6c\ud604\ud558\uae30 \uc704\ud574 \uc0dd\uc131\uc790\uc640 \ud310\ubcc4\uc790 \ubaa8\ub378\uc744 \uc815\uc758\ud569\ub2c8\ub2e4.<\/p>\n 3.1 \uc0dd\uc131\uc790 \ubaa8\ub378<\/h3>\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, 28 * 28), # MNIST \uc774\ubbf8\uc9c0 \ud06c\uae30\n nn.Tanh() # \uc0dd\uc131\ub41c \uc774\ubbf8\uc9c0\uc758 \ud53d\uc140 \uac12 \ubc94\uc704\ub97c -1 ~ 1\ub85c \uc870\uc815\n )\n\n def forward(self, z):\n img = self.model(z)\n img = img.view(img.size(0), 1, 28, 28) # \uc774\ubbf8\uc9c0 \ud615\ud0dc\ub85c \ubcc0\ud658\n return img<\/code><\/pre>\n3.2 \ud310\ubcc4\uc790 \ubaa8\ub378<\/h3>\nclass Discriminator(nn.Module):\n def __init__(self):\n super(Discriminator, self).__init__()\n self.model = nn.Sequential(\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() # \uc2e4\uc218\uac00 \uc544\ub2cc 0\uacfc 1 \uc0ac\uc774\uc758 \uac12\uc73c\ub85c \ucd9c\ub825\n )\n\n def forward(self, img):\n img_flat = img.view(img.size(0), -1) # \uc774\ubbf8\uc9c0\ub97c \ud3c9\ud0c4\ud654\n validity = self.model(img_flat)\n return validity<\/code><\/pre>\n4. GAN\uc758 \ud559\uc2b5 \uacfc\uc815<\/h2>\nGAN\uc758 \ud559\uc2b5 \uacfc\uc815\uc740 \ub2e4\uc74c\uacfc \uac19\uc774 \uc774\ub8e8\uc5b4\uc9d1\ub2c8\ub2e4.<\/p>\n \n\uc9c4\uc9dc \uc774\ubbf8\uc9c0\uc640 \uc0dd\uc131\ub41c \uc774\ubbf8\uc9c0\ub97c \ud310\ubcc4\uc790\uc5d0 \uc81c\uacf5\ud558\uc5ec \ud310\ubcc4\uc790\uc758 \uc190\uc2e4(loss)\uc744 \uacc4\uc0b0\ud569\ub2c8\ub2e4.<\/li>\n \uc0dd\uc131\uc790\ub97c \uc5c5\ub370\uc774\ud2b8\ud558\uc5ec \uc0dd\uc131\ub41c \uc774\ubbf8\uc9c0\uac00 \uc2e4\uc81c\uc640 \uac00\uae4c\uc6cc\uc9c0\ub3c4\ub85d \ud569\ub2c8\ub2e4.<\/li>\n \uc774 \uacfc\uc815\uc744 \ubc18\ubcf5\ud558\uc5ec \uac01 \ub124\ud2b8\uc6cc\ud06c\uac00 \uc11c\ub85c \ubc1c\uc804\ud558\uac8c \ud569\ub2c8\ub2e4.<\/li>\n<\/ul>\n4.1 \uc190\uc2e4 \ud568\uc218 \ubc0f \ucd5c\uc801\ud654\uae30 \uc815\uc758<\/h3>\nimport torch.optim as optim\n\n# \uc0dd\uc131\uc790 \ubc0f \ud310\ubcc4\uc790 \uc778\uc2a4\ud134\uc2a4 \uc0dd\uc131\ngenerator = Generator()\ndiscriminator = Discriminator()\n\n# \uc190\uc2e4 \ud568\uc218 \ubc0f \ucd5c\uc801\ud654\uae30 \uc124\uc815\nadversarial_loss = 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))<\/code><\/pre>\n4.2 \ud559\uc2b5 \ub8e8\ud504<\/h3>\nnum_epochs = 200\nfor epoch in range(num_epochs):\n for i, (imgs, _) in enumerate(data_loader):\n # \uc2e4\uc81c \ub370\uc774\ud130\uc758 \ub808\uc774\ube14: 1\n real_imgs = imgs\n valid = torch.ones(imgs.size(0), 1) # \uc9c4\uc9dc \uc774\ubbf8\uc9c0\uc5d0 \ub300\ud55c \uc815\ub2f5\n fake = torch.zeros(imgs.size(0), 1) # \uac00\uc9dc \uc774\ubbf8\uc9c0\uc5d0 \ub300\ud55c \uc815\ub2f5\n\n # \ud310\ubcc4\uc790 \ud559\uc2b5\n optimizer_D.zero_grad()\n z = torch.randn(imgs.size(0), 100) # \ubb34\uc791\uc704 \ub178\uc774\uc988 \uc0d8\ud50c\ub9c1\n generated_imgs = generator(z) # \uc0dd\uc131\ub41c \uc774\ubbf8\uc9c0\n real_loss = adversarial_loss(discriminator(real_imgs), valid)\n fake_loss = adversarial_loss(discriminator(generated_imgs.detach()), fake)\n d_loss = (real_loss + fake_loss) \/ 2\n d_loss.backward()\n optimizer_D.step()\n\n # \uc0dd\uc131\uc790 \ud559\uc2b5\n optimizer_G.zero_grad()\n g_loss = adversarial_loss(discriminator(generated_imgs), valid)\n g_loss.backward()\n optimizer_G.step()\n\n print(f'Epoch {epoch}\/{num_epochs} | D Loss: {d_loss.item()} | G Loss: {g_loss.item()}')<\/code><\/pre>\n5. \uacb0\uacfc \uc2dc\uac01\ud654<\/h2>\n\ud559\uc2b5\uc774 \uc644\ub8cc\ub41c \ud6c4\uc5d0 \uc0dd\uc131\ub41c \uc774\ubbf8\uc9c0\ub97c \uc2dc\uac01\ud654\ud558\uc5ec \uacb0\uacfc\ub97c \ud655\uc778\ud569\ub2c8\ub2e4.<\/p>\n import matplotlib.pyplot as plt\n\n# \uc0dd\uc131\ub41c \uc774\ubbf8\uc9c0 \uc2dc\uac01\ud654\ndef show_generated_images(generator, num_images=16):\n z = torch.randn(num_images, 100) # \ubb34\uc791\uc704 \ub178\uc774\uc988 \uc0d8\ud50c\ub9c1\n generated_images = generator(z)\n generated_images = generated_images.detach().numpy()\n\n fig, axs = plt.subplots(4, 4, figsize=(10, 10))\n for i in range(4):\n for j in range(4):\n axs[i, j].imshow(generated_images[i * 4 + j, 0], cmap='gray')\n axs[i, j].axis('off')\n plt.show()\n\nshow_generated_images(generator)<\/code><\/pre>\n\uc774\uc640 \uac19\uc740 \ubc29\ubc95\uc73c\ub85c GAN\uc744 \uad6c\ucd95\ud558\uace0 \ud559\uc2b5\uc2dc\ucf1c \uc0dd\uc131\ub41c \uc774\ubbf8\uc9c0\ub97c \ud655\uc778\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. GAN\uc758 \ud65c\uc6a9 \uac00\ub2a5\uc131\uc740 \ub9e4\uc6b0 \ub113\uc73c\uba70, \uc774\ub97c \ud1b5\ud574 \ucc3d\uc758\uc801\uc778 \uc791\uc5c5\uc744 \uc218\ud589\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc774\uc81c \uc5ec\ub7ec\ubd84\ub3c4 GAN\uc758 \uc138\uacc4\uc5d0 \ud55c \uac78\uc74c \ub354 \uac00\uae4c\uc6cc\uc9c8 \uc218 \uc788\uc2b5\ub2c8\ub2e4!<\/p>\n 6. \uacb0\ub860<\/h2>\nGenerative Adversarial Networks\ub294 \ub9e4\uc6b0 \ud765\ubbf8\ub85c\uc6b4 \ub525\ub7ec\ub2dd \ubd84\uc57c\ub85c, \ub9ce\uc740 \uc5f0\uad6c\uc640 \uac1c\ubc1c\uc5d0\uc11c \ud65c\ubc1c\ud788 \uc0ac\uc6a9\ub418\uace0 \uc788\uc2b5\ub2c8\ub2e4. \uc774\ubc88 \uac15\uc88c\uc5d0\uc11c\ub294 \ud30c\uc774\ud1a0\uce58\ub97c \ud65c\uc6a9\ud558\uc5ec GAN\uc758 \uae30\ubcf8 \uc6d0\ub9ac\uc640 \uad6c\uc870\ub97c \uc0b4\ud3b4\ubcf4\uc558\uace0, \uc2e4\uc81c\ub85c \ub525\ub7ec\ub2dd \ubaa8\ub378\uc744 \uad6c\ucd95\ud558\uace0 \ud559\uc2b5\ud558\ub294 \uacfc\uc815\uc744 \ub2e4\ub8e8\uc5c8\uc2b5\ub2c8\ub2e4. \uc5ec\ub7ec\ubd84\uc774 \uc774 \uac15\uc88c\ub97c \ud1b5\ud574 GAN\uc5d0 \ub300\ud55c \uae4a\uc740 \uc774\ud574\uc640 \ud765\ubbf8\ub97c \ub290\ub07c\uae38 \ubc14\ub77c\uba70, \uc55e\uc73c\ub85c\uc758 \ub525\ub7ec\ub2dd \uc5ec\uc815\uc5d0 \ud070 \ub3c4\uc6c0\uc774 \ub418\uae38 \ubc14\ub78d\ub2c8\ub2e4.<\/p>\n