{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python Docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load\n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n\n# Input data files are available in the read-only \"../input/\" directory\n# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory\n\nimport os\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n    for filename in filenames:\n        print(os.path.join(dirname, filename))\n\n# You can write up to 20GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using \"Save & Run All\" \n# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import torch\nimport numpy as npy\nfrom torch.utils.data import Dataset,DataLoader\nimport torch.nn as nn\nimport torch.nn.functional as F\nimport os\nfrom torchvision.transforms import transforms\nimport zipfile\nimport matplotlib.pyplot as plt\nfrom PIL import Image","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"with zipfile.ZipFile('/kaggle/input/dogs-vs-cats/train.zip','r') as z:\n    z.extractall()\nwith zipfile.ZipFile('/kaggle/input/dogs-vs-cats/test1.zip','r') as z:\n    z.extractall()","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"class dataset(Dataset):\n    def __init__(self,path,trans):\n        self.path=path\n        self.trans=trans\n        self.img_path=os.listdir(self.path)\n    def __getitem__(self,index):\n        name=self.img_path[index]\n        print(name)\n        img=Image.open(self.path+'/'+name)\n        now=name.split('.')[0]\n        label=0\n        if now=='dog':\n            label=1\n        return self.trans(img),label\n    def __len__(self):\n        return len(self.img_path)\n\n    \ntrans=transforms.Compose([transforms.Resize((224,224)),transforms.ToTensor(),transforms.Normalize((0.5,0.5,0.5),(0.5,0.5,0.5))])\ntrain_set=dataset('/kaggle/working/train',trans)\ntest_set=dataset('/kaggle/working/test1',trans)\ntrain_loader=DataLoader(dataset=train_set,batch_size=4,shuffle=True)\ntest_loader=DataLoader(dataset=test_set,batch_size=1,shuffle=False)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"f=open('./show.txt','w+')\nclass VGG16(nn.Module):\n    def __init__(self):\n        super(VGG16,self).__init__()\n        #3*224*224\n        self.conv1_1=nn.Conv2d(3,64,3)\n        self.b1_1=nn.BatchNorm2d(64)\n        # 64*222*222\n        self.conv1_2=nn.Conv2d(64,64,3,padding=(1,1))\n        self.b1_2=nn.BatchNorm2d(64)\n        #64*222*222\n        self.maxpool1=nn.MaxPool2d((2,2),padding=(1,1))\n        #64*112*112\n        \n        self.conv2_1=nn.Conv2d(64,128,3)\n        self.b2_1=nn.BatchNorm2d(128)\n        #128*110*110\n        self.conv2_2=nn.Conv2d(128,128,3,padding=(1,1))\n        self.b2_2=nn.BatchNorm2d(128)\n        #128*110*110\n        self.maxpool2=nn.MaxPool2d((2,2),padding=(1,1))\n        #128*56*56\n        \n        self.conv3_1=nn.Conv2d(128,256,3)\n        self.b3_1=nn.BatchNorm2d(256)\n        #256*54*54\n        self.conv3_2=nn.Conv2d(256,256,3,padding=(1,1))\n        self.b3_2=nn.BatchNorm2d(256)\n        #256*54*54\n        self.conv3_3=nn.Conv2d(256,256,3,padding=(1,1))\n        self.b3_3=nn.BatchNorm2d(256)\n        #256*54*54\n        self.maxpool3=nn.MaxPool2d((2,2),padding=(1,1))\n        #256*28*28\n        \n        self.conv4_1=nn.Conv2d(256,512,3)\n        self.b4_1=nn.BatchNorm2d(512)\n        #512*26*26\n        self.conv4_2=nn.Conv2d(512,512,3,padding=(1,1))\n        self.b4_2=nn.BatchNorm2d(512)\n        #512*26*26\n        self.conv4_3=nn.Conv2d(512,512,3,padding=(1,1))\n        self.b4_3=nn.BatchNorm2d(512)\n        #512*26*26\n        self.maxpool4=nn.MaxPool2d((2,2),padding=(1,1))\n        #512*14*14\n        \n        self.conv5_1=nn.Conv2d(512,512,3)\n        self.b5_1=nn.BatchNorm2d(512)\n        #512*12*12\n        self.conv5_2=nn.Conv2d(512,512,3,padding=(1,1))\n        self.b5_2=nn.BatchNorm2d(512)\n        #512*12*12\n        self.conv5_3=nn.Conv2d(512,512,3,padding=(1,1))\n        self.b5_3=nn.BatchNorm2d(512)\n        #512*12*12\n        self.maxpool5=nn.MaxPool2d((2,2),padding=(1,1))\n        #512*7*7\n        \n        self.f1c=nn.Linear(512*7*7,4096)\n        self.f2c=nn.Linear(4096,4096)\n        self.f3c=nn.Linear(4096,2)\n        \n    def forward(self, x):\n        # x.size(0)即为batch_size\n        in_size = x.size(0)\n        out = self.conv1_1(x)  # 222\n        out = self.b1_1(out)\n        out = F.relu(out)\n        out = self.conv1_2(out)  # 222\n        out = self.b1_2(out)\n        out = F.relu(out)\n        out = self.maxpool1(out)  # 112\n\n        out = self.conv2_1(out)  # 110\n        out = self.b2_1(out)\n        out = F.relu(out)\n        out = self.conv2_2(out)  # 110\n        out = self.b2_2(out)\n        out = F.relu(out)\n        out = self.maxpool2(out)  # 56\n\n        out = self.conv3_1(out)  # 54\n        out = self.b3_1(out)\n        out = F.relu(out)\n        out = self.conv3_2(out)  # 54\n        out = self.b3_2(out)\n        out = F.relu(out)\n        out = self.conv3_3(out)  # 54\n        out = self.b3_3(out)\n        out = F.relu(out)\n        out = self.maxpool3(out)  # 28\n\n\n        \n        \n        out = self.conv4_1(out)  # 26\n        out = self.b4_1(out)\n        out = F.relu(out)\n\n        out = self.conv4_2(out)  # 26\n        out = self.b4_2(out)\n        out = F.relu(out)\n        out = self.conv4_3(out)  # 26\n        out = self.b4_3(out)\n        out = F.relu(out)\n        out = self.maxpool4(out)  # 14\n        \n        out = self.conv5_1(out)  # 12\n        out = self.b5_1(out)\n        out = F.relu(out)\n        out = self.conv5_2(out)  # 12\n        out = self.b5_2(out)\n        out = F.relu(out)\n        out = self.conv5_3(out)  # 12\n        out = self.b5_3(out)\n        out = F.relu(out)\n        out = self.maxpool5(out)  # 7\n        \n        # 展平\n        out = out.view(in_size, -1)\n        out = self.f1c(out)\n        out = F.relu(out)\n        out = self.f2c(out)\n        out = F.relu(out)\n        out = self.f3c(out)\n\n        out = F.softmax(out, dim=1)\n\n        return out\nnet=VGG16()\nif torch.cuda.is_available():\n    net=net.cuda()\nfor name in net.state_dict():\n    print(name)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"loss=nn.CrossEntropyLoss()\nif torch.cuda.is_available():\n    loss=loss.cuda()\n\noptim=torch.optim.SGD(net.parameters(),lr=0.0001)\nnet.train()\nfor epoch in range(5):\n    for data in train_loader:\n        imgs,labels=data\n        if torch.cuda.is_available():\n            imgs=imgs.cuda()\n            labels=labels.cuda()\n        outputs=net(imgs)\n        print(outputs)\n        result_loss=loss(outputs,labels)\n        optim.zero_grad()\n        result_loss.backward()\n        optim.step()\n    print(result_loss)\n\n\nf.close()\n    ","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"net.eval()\nans=[]\nfor data in test_loader:\n    img,label=data\n    print(img)\n    if torch.cuda.is_available():\n        img=img.cuda()\n    output=net(img)\n    print(output)\n    if output[0][1]>output[0][0]:\n        ans.append('dog')\n    else:\n        ans.append('cat')\nans=npy.array(ans)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"submission=pd.read_csv(\"../input/dogs-vs-cats/sampleSubmission.csv\")\nsubmission.label = ans\nsubmission.to_csv('predictions.csv', index=False)","metadata":{"trusted":true},"execution_count":null,"outputs":[]}]}