cnn_main.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "import cv2 as cv\n",
    "import numpy as np\n",
    "import torch\n",
    "import torch.nn.functional as F\n",
    "import torch.utils.data as data\n",
    "import torchvision.transforms as transforms\n",
    "from matplotlib import pyplot as plt\n",
    "from PIL import Image\n",
    "from torch import nn, optim\n",
    "from torch.utils.data import DataLoader\n",
    "from my_models import *\n",
    "from cat_dog_dataset import my_Dataset\n",
    "global PATH, IMG_SIZE, BATCH_SIZE, DEVICE\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "def train(epoch, model, train_dataloader, criterion, optimizer):\n",
    "    '''模型训练函数'''\n",
    "    global print_loss\n",
    "    # 调整为训练模式\n",
    "    model.train()\n",
    "    losses = []\n",
    "    plot_loss = 0\n",
    "    total_num = len(train_dataloader.dataset)\n",
    "    print_thre = total_num//(BATCH_SIZE*5)\n",
    "    for i, (img, label) in enumerate(train_dataloader):\n",
    "        img, label = img.to(DEVICE), label.to(DEVICE)\n",
    "        out = model(img)\n",
    "        # 计算损失\n",
    "        loss = criterion(out, label.squeeze())\n",
    "        losses.append(loss.item())\n",
    "        # 误差反向传播\n",
    "        loss.backward()\n",
    "        # 参数调整\n",
    "        optimizer.step()\n",
    "        # 清除优化器中梯度\n",
    "        optimizer.zero_grad()\n",
    "        if (i+1) % print_thre == 0:\n",
    "            plot_loss += np.mean(losses)\n",
    "            print('TRAINING: Epoch:{} [{} / {} ({:.2f}%)] Loss:{}'.format(\n",
    "                epoch, (i+1)*BATCH_SIZE, total_num, (i+1)*BATCH_SIZE/total_num*100, plot_loss))\n",
    "            losses.clear()\n",
    "    print_loss.append(plot_loss/5)\n",
    "\n",
    "\n",
    "def test(model, test_dataloader, criterion):\n",
    "    '''模型验证函数'''\n",
    "    model.eval()\n",
    "    cnt = 0\n",
    "    losses = []\n",
    "    total_num = len(test_dataloader.dataset)\n",
    "    for i, (img, label) in enumerate(test_dataloader):\n",
    "        img, label = img.to(DEVICE), label.to(DEVICE)\n",
    "        out = model(img)\n",
    "        # 计算损失\n",
    "        loss = criterion(out, label.squeeze())\n",
    "        losses.append(loss.item())\n",
    "        y = out.argmax(1, keepdim=True)\n",
    "        num = (y-label).abs().sum()\n",
    "        cnt += num\n",
    "    acc = 1-cnt/total_num\n",
    "    print('\\nTESTING: Loss:{},Accuracy:[{} / {} ({:.6f}%)]\\n'.format(np.mean(losses), total_num-cnt, total_num,\n",
    "          acc*100))\n",
    "    return acc.item()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "using device cuda\n"
     ]
    }
   ],
   "source": [
    "global PATH, IMG_SIZE, BATCH_SIZE, DEVICE\n",
    "'''定义相关超参数'''\n",
    "# 训练集路径\n",
    "PATH = \"data/train\"\n",
    "# 裁剪图像大小\n",
    "IMG_SIZE = 256\n",
    "BATCH_SIZE = 25\n",
    "EPOCHS = 10\n",
    "DEVICE = torch.device(\n",
    "    \"cuda\") if torch.cuda.is_available() else torch.device(\"cpu\")\n",
    "WORKERS = 0\n",
    "# 训练数据比例\n",
    "TRAIN_RATE = 0.7\n",
    "\n",
    "print(\"using device {}\".format(DEVICE))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 数据集分割，采用固定的seed\n",
    "img_data = my_Dataset(\"train\", PATH, IMG_SIZE)\n",
    "IMG_NUM = len(img_data)\n",
    "TRAIN_NUM = int(IMG_NUM*TRAIN_RATE)\n",
    "train_data, vali_data = data.random_split(\n",
    "    img_data, [TRAIN_NUM, IMG_NUM-TRAIN_NUM], torch.Generator().manual_seed(1))\n",
    "train_dataloader = DataLoader(train_data, BATCH_SIZE,\n",
    "                                shuffle=True, num_workers=WORKERS, drop_last=True)\n",
    "vali_dataloader = DataLoader(vali_data, BATCH_SIZE,\n",
    "                                shuffle=True, num_workers=WORKERS, drop_last=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "using model: Net\n"
     ]
    }
   ],
   "source": [
    "# 定义网络模型\n",
    "model = get_cnn_model(True).to(DEVICE)\n",
    "# model = get_resnet18_model(False, True).to(DEVICE)\n",
    "# model = get_vgg16_model(True, False).to(DEVICE)\n",
    "print(\"using model:\", model._get_name())\n",
    "optimizer = optim.Adam(model.parameters(),lr=0.0001)\n",
    "criterion = nn.CrossEntropyLoss()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "print_loss = []\n",
    "accs = []\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TRAINING: Epoch:0 [3500 / 17500 (20.00%)] Loss:0.4404014434133257\n",
      "TRAINING: Epoch:0 [7000 / 17500 (40.00%)] Loss:0.8797682538628577\n",
      "TRAINING: Epoch:0 [10500 / 17500 (60.00%)] Loss:1.3218474905405726\n",
      "TRAINING: Epoch:0 [14000 / 17500 (80.00%)] Loss:1.7658033888254847\n",
      "TRAINING: Epoch:0 [17500 / 17500 (100.00%)] Loss:2.206483416046415\n",
      "\n",
      "TESTING: Loss:0.5086086635788282,Accuracy:[5944 / 7500 (79.253334%)]\n",
      "\n",
      "TRAINING: Epoch:1 [3500 / 17500 (20.00%)] Loss:0.4406935977084296\n",
      "TRAINING: Epoch:1 [7000 / 17500 (40.00%)] Loss:0.8765077995402473\n",
      "TRAINING: Epoch:1 [10500 / 17500 (60.00%)] Loss:1.3239361858793668\n",
      "TRAINING: Epoch:1 [14000 / 17500 (80.00%)] Loss:1.757505672318595\n",
      "TRAINING: Epoch:1 [17500 / 17500 (100.00%)] Loss:2.190777201524803\n",
      "\n",
      "TESTING: Loss:0.5098976649840673,Accuracy:[5956 / 7500 (79.413330%)]\n",
      "\n",
      "TRAINING: Epoch:2 [3500 / 17500 (20.00%)] Loss:0.4273797765374184\n",
      "TRAINING: Epoch:2 [7000 / 17500 (40.00%)] Loss:0.8643233058708055\n",
      "TRAINING: Epoch:2 [10500 / 17500 (60.00%)] Loss:1.2974197213138854\n",
      "TRAINING: Epoch:2 [14000 / 17500 (80.00%)] Loss:1.7456735634378027\n",
      "TRAINING: Epoch:2 [17500 / 17500 (100.00%)] Loss:2.181323457189969\n",
      "\n",
      "TESTING: Loss:0.5105984732508659,Accuracy:[5943 / 7500 (79.239998%)]\n",
      "\n",
      "TRAINING: Epoch:3 [3500 / 17500 (20.00%)] Loss:0.4352688065596989\n",
      "TRAINING: Epoch:3 [7000 / 17500 (40.00%)] Loss:0.8663798044834818\n",
      "TRAINING: Epoch:3 [10500 / 17500 (60.00%)] Loss:1.2947438225150107\n",
      "TRAINING: Epoch:3 [14000 / 17500 (80.00%)] Loss:1.7293738795178277\n",
      "TRAINING: Epoch:3 [17500 / 17500 (100.00%)] Loss:2.161771833896637\n",
      "\n",
      "TESTING: Loss:0.508514275153478,Accuracy:[5955 / 7500 (79.400002%)]\n",
      "\n",
      "TRAINING: Epoch:4 [3500 / 17500 (20.00%)] Loss:0.424084495646613\n",
      "TRAINING: Epoch:4 [7000 / 17500 (40.00%)] Loss:0.8516035422682762\n",
      "TRAINING: Epoch:4 [10500 / 17500 (60.00%)] Loss:1.2854793078133038\n",
      "TRAINING: Epoch:4 [14000 / 17500 (80.00%)] Loss:1.7198353677988052\n",
      "TRAINING: Epoch:4 [17500 / 17500 (100.00%)] Loss:2.1489569385136877\n",
      "\n",
      "TESTING: Loss:0.5064724566539128,Accuracy:[5973 / 7500 (79.639999%)]\n",
      "\n",
      "TRAINING: Epoch:5 [3500 / 17500 (20.00%)] Loss:0.4232906147837639\n",
      "TRAINING: Epoch:5 [7000 / 17500 (40.00%)] Loss:0.8453137493559293\n",
      "TRAINING: Epoch:5 [10500 / 17500 (60.00%)] Loss:1.2769785165786744\n",
      "TRAINING: Epoch:5 [14000 / 17500 (80.00%)] Loss:1.7103596150875093\n",
      "TRAINING: Epoch:5 [17500 / 17500 (100.00%)] Loss:2.1394538457904546\n",
      "\n",
      "TESTING: Loss:0.5081452090541522,Accuracy:[5973 / 7500 (79.639999%)]\n",
      "\n",
      "TRAINING: Epoch:6 [3500 / 17500 (20.00%)] Loss:0.4226202204823494\n",
      "TRAINING: Epoch:6 [7000 / 17500 (40.00%)] Loss:0.8488567320363862\n",
      "TRAINING: Epoch:6 [10500 / 17500 (60.00%)] Loss:1.2780507294195038\n",
      "TRAINING: Epoch:6 [14000 / 17500 (80.00%)] Loss:1.7047370814851353\n",
      "TRAINING: Epoch:6 [17500 / 17500 (100.00%)] Loss:2.1227820566722326\n",
      "\n",
      "TESTING: Loss:0.5036147885521253,Accuracy:[6000 / 7500 (80.000000%)]\n",
      "\n",
      "TRAINING: Epoch:7 [3500 / 17500 (20.00%)] Loss:0.42205136993101666\n",
      "TRAINING: Epoch:7 [7000 / 17500 (40.00%)] Loss:0.8399485030344555\n",
      "TRAINING: Epoch:7 [10500 / 17500 (60.00%)] Loss:1.262935962847301\n",
      "TRAINING: Epoch:7 [14000 / 17500 (80.00%)] Loss:1.6931066006422042\n",
      "TRAINING: Epoch:7 [17500 / 17500 (100.00%)] Loss:2.1147936708160806\n",
      "\n",
      "TESTING: Loss:0.5072244941194852,Accuracy:[5956 / 7500 (79.413330%)]\n",
      "\n",
      "TRAINING: Epoch:8 [3500 / 17500 (20.00%)] Loss:0.41812770515680314\n",
      "TRAINING: Epoch:8 [7000 / 17500 (40.00%)] Loss:0.8318883129528591\n",
      "TRAINING: Epoch:8 [10500 / 17500 (60.00%)] Loss:1.2533649099724635\n",
      "TRAINING: Epoch:8 [14000 / 17500 (80.00%)] Loss:1.677289351820946\n",
      "TRAINING: Epoch:8 [17500 / 17500 (100.00%)] Loss:2.0998857572674754\n",
      "\n",
      "TESTING: Loss:0.5041355086366336,Accuracy:[5994 / 7500 (79.919998%)]\n",
      "\n",
      "TRAINING: Epoch:9 [3500 / 17500 (20.00%)] Loss:0.41056180234466283\n",
      "TRAINING: Epoch:9 [7000 / 17500 (40.00%)] Loss:0.8299495492662703\n",
      "TRAINING: Epoch:9 [10500 / 17500 (60.00%)] Loss:1.2515160677688464\n",
      "TRAINING: Epoch:9 [14000 / 17500 (80.00%)] Loss:1.6737901049000878\n",
      "TRAINING: Epoch:9 [17500 / 17500 (100.00%)] Loss:2.0875109738537247\n",
      "\n",
      "TESTING: Loss:0.5085571520527203,Accuracy:[5949 / 7500 (79.320000%)]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "for epoch in range(EPOCHS):\n",
    "    train(epoch, model, train_dataloader, criterion, optimizer)\n",
    "    if (epoch+1) % 1 == 0:\n",
    "        acc = test(model, vali_dataloader, criterion)\n",
    "        accs.append(acc)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "模型参数保存成功\n"
     ]
    }
   ],
   "source": [
    "if acc >= 0.75:\n",
    "    torch.save(model.state_dict(), \"cnn3_state.pth\")\n",
    "    print(\"模型参数保存成功\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(1)\n",
    "plt.plot(range(EPOCHS), print_loss,   linestyle='-.', linewidth=2,\n",
    "         marker='o', markersize=8, markeredgecolor='k', markerfacecolor='red')\n",
    "plt.xlabel(\"EPOCHS\")\n",
    "plt.ylabel(\"Loss\")\n",
    "plt.title(\"CNN Loss下降曲线\")\n",
    "plt.savefig(\"Loss_cnn.jpg\")\n",
    "\n",
    "plt.figure(2)\n",
    "plt.plot(range(EPOCHS), accs,linestyle='-.', linewidth=2,\n",
    "         marker='o', markersize=8, markeredgecolor='k', markerfacecolor='red')\n",
    "plt.xlabel(\"EPOCHS\")\n",
    "plt.ylabel(\"Accuracy\")\n",
    "plt.title(\"CNN 准确率变化情况\")\n",
    "plt.savefig(\"Accuracy_cnn.jpg\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.1 (tags/v3.9.1:1e5d33e, Dec  7 2020, 17:08:21) [MSC v.1927 64 bit (AMD64)]"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "11938c6bc6919ae2720b4d5011047913343b08a43b18698fd82dedb0d4417594"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}