DCGAN.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "https://mp.weixin.qq.com/s/CtFvURdJR314OBFDhu5hGg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 导入所需的库\n",
    "import torch\n",
    "from torch import nn\n",
    "from torch import optim\n",
    "from torchvision.transforms import transforms\n",
    "import numpy as np\n",
    "from PIL import Image\n",
    "import matplotlib.pyplot as plt\n",
    "from tqdm.notebook import tqdm\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([3, 32, 32])\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "transform = transforms.Compose(\n",
    "    [\n",
    "        transforms.Resize((32,32)),\n",
    "        transforms.ToTensor()\n",
    "    ]\n",
    ")\n",
    "flat_img = 3072\n",
    "img = Image.open('../dataset/truck.jpg')\n",
    "real_img = transform(img)\n",
    "torch.manual_seed(2)\n",
    "fake_img = torch.rand(1,100)\n",
    "# print(real_img.size())\n",
    "plt.imshow(np.transpose(real_img.numpy(),(1,2,0)))\n",
    "# plt.imshow(real_img.numpy())\n",
    "print(real_img.size())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 创建一个判别器模型\n",
    "class Discriminator(nn.Module):\n",
    "    def __init__(self):\n",
    "        super().__init__()\n",
    "        self.linear = nn.Sequential(\n",
    "            nn.Linear(flat_img,10000),\n",
    "            nn.ReLU(),\n",
    "            nn.Linear(10000,1),\n",
    "            nn.Sigmoid()\n",
    "        )\n",
    "    def forward(self,img):\n",
    "        img = img.view(1,-1)\n",
    "        out = self.linear(img)\n",
    "        return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 创建一个生成器模型\n",
    "class Generator(nn.Module):\n",
    "    def __init__(self):\n",
    "        super().__init__()\n",
    "        self.linear = nn.Sequential(\n",
    "            nn.Linear(100,10000),\n",
    "            nn.LeakyReLU(),\n",
    "            nn.Linear(10000,4000),\n",
    "            nn.LeakyReLU(),\n",
    "            nn.Linear(4000,flat_img)\n",
    "        )\n",
    "    def forward(self,latent_space):\n",
    "        latent_space = latent_space.view(1,-1)\n",
    "        out = self.linear(latent_space)\n",
    "        return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 初始化模型，优化器和损失函数\n",
    "device = 'cuda:0' if torch.cuda.is_available() else 'cpu'\n",
    "discr = Discriminator().to(device)\n",
    "gen = Generator().to(device)\n",
    "opt_d = optim.SGD(discr.parameters(),lr=0.001,momentum=0.9)\n",
    "opt_g = optim.SGD(gen.parameters(),lr=0.001,momentum=0.9)\n",
    "criterion = nn.BCELoss()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "epochs = 500\n",
    "discr_e = 4\n",
    "gen_e = 3\n",
    "for epoch in tqdm(range(epochs),total=epochs):\n",
    "    for k in range(discr_e):\n",
    "        opt_d.zero_grad()\n",
    "        out_d1 = discr(real_img.to(device))\n",
    "        # loss for real image\n",
    "        loss_d1 = criterion(out_d1,torch.ones((1,1)).to(device))\n",
    "        loss_d1.backward()\n",
    "        out_d2 = gen(fake_img.to(device)).detach()\n",
    "        # loss for fake image\n",
    "        loss_d2 = criterion(discr(out_d2.to(device)),torch.zeros((1,1)).to(device))\n",
    "        loss_d2.backward()\n",
    "        opt_d.step()\n",
    "    # generator training\n",
    "    for i in range(gen_e):\n",
    "        opt_g.zero_grad()\n",
    "        out_g = gen(fake_img.to(device))\n",
    "        loss_g = torch.log(1.0-(discr(out_g.to(device))))\n",
    "        loss_g.backward()\n",
    "        opt_g.step()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:root]",
   "language": "python",
   "name": "conda-root-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}