{"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\nfrom torch import nn\nfrom torch.utils.data import Dataset, DataLoader, random_split\nimport torch.nn.functional as F\n\nimport torchvision\nfrom torchvision import transforms, datasets, models\nfrom torchvision.datasets import ImageFolder\n\nfrom PIL import Image\nimport matplotlib.pyplot as plt","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Image directory building","metadata":{}},{"cell_type":"code","source":"base_dir = \"../input/dogs-vs-cats\"\ntrain_zip = os.path.join(base_dir, \"train.zip\")\ntest_zip = os.path.join(base_dir, \"test1.zip\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Unzip datasets\nimport zipfile\nwith zipfile.ZipFile(train_zip,\"r\") as z:\n    z.extractall()\nwith zipfile.ZipFile(test_zip, \"r\") as z:\n    z.extractall()","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_dir = \"./train\"","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Make directorys which save data (For ImageFolder)\ntry:\n    if not os.path.exists(f\"{train_dir}/cat\"):\n        os.makedirs(f\"{train_dir}/cat\")\n    if not os.path.exists(f\"{train_dir}/dog\"):\n        os.makedirs(f\"{train_dir}/dog\")\nexcept OSError:\n    print(\"ERROR : Creating Directory\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import shutil\n\n# Move data according to its name\nfor file_name in os.listdir(train_dir):\n    try:\n        cat_or_dog, _, _ = file_name.split(\".\")\n        if cat_or_dog == \"cat\":\n            shutil.move(f\"{train_dir}/{file_name}\", f\"{train_dir}/cat/{file_name}\")\n        elif cat_or_dog == \"dog\":\n            shutil.move(f\"{train_dir}/{file_name}\", f\"{train_dir}/dog/{file_name}\")\n        else:\n            print(\"ERROR : No label\")\n            break\n    except:\n        continue\n        \nprint(os.listdir(train_dir))","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Make Custom DataLoader via ImageFolder","metadata":{}},{"cell_type":"code","source":"data_transform = transforms.Compose([\n    transforms.Resize((256, 256)),\n    transforms.ToTensor(),\n    transforms.Normalize((0.5,0.5,0.5), (0.5,0.5,0.5))\n])\n\ndataset = torchvision.datasets.ImageFolder(root=train_dir, transform=data_transform)\nprint(dataset.class_to_idx)\nidx_to_class = {v:k for k,v in dataset.class_to_idx.items()}","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"data_size = len(dataset)\n# Random split dataset\ntrain_dataset, validation_dataset = random_split(dataset, [int(0.8 * data_size), int(0.2 * data_size)])\n\ntrain_dataloader = DataLoader(train_dataset, batch_size=1, shuffle=True)\nvalidation_dataloader = DataLoader(validation_dataset, batch_size=1, shuffle=True)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"for value in train_dataloader:\n    data, label = value\n    print(data.shape)\n    print(f\"This is... {idx_to_class[int(label.item())]}\")\n    plt.imshow(torch.squeeze(data).permute(1, 2, 0))\n    break","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Make Prediction Model","metadata":{}},{"cell_type":"code","source":"device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\nprint(f\"Using {device} device\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Scratch traing","metadata":{}},{"cell_type":"code","source":"# Accuraccy 50% model == Equal to random choice\nclass VanilaMyNet(nn.Module):\n    def __init__(self):\n        super(VanilaMyNet, self).__init__()\n\n        self.conv1 = nn.Conv2d(in_channels=3, out_channels=12, kernel_size=5, stride=1, padding=1)\n        self.conv2 = nn.Conv2d(in_channels=12, out_channels=24, kernel_size=3, stride=1, padding=1)\n        self.bn1 = nn.BatchNorm2d(24)\n        self.pool = nn.MaxPool2d(2,2)\n        self.conv3 = nn.Conv2d(in_channels=24, out_channels=12, kernel_size=5, stride=1, padding=1)\n        self.bn2 = nn.BatchNorm2d(12)\n        self.fc1 = nn.Linear(187500, 2)\n\n    def forward(self, input):\n        output = F.relu(self.bn1(self.conv2(self.conv1(input))))\n        output = self.pool(output)\n        output = F.relu(self.bn2(self.conv3(output)))\n        output = output.view(-1, 187500)\n        output = self.fc1(output)\n\n        return output","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Transfer learning","metadata":{}},{"cell_type":"code","source":"resnet18_pretrained = models.resnet18(pretrained=True,progress=True)\nnum_ftrs = resnet18_pretrained.fc.in_features\nnum_classes = 2\nresnet18_pretrained.fc = nn.Linear(num_ftrs, num_classes)\nprint(resnet18_pretrained)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Train Model","metadata":{}},{"cell_type":"code","source":"def train(dataloader, model, loss_fn, optimizer):\n    size = len(dataloader.dataset)\n    for batch, (X, y) in enumerate(dataloader):\n        X, y = X.to(device), y.to(device)\n\n        pred = model(X)\n        loss = loss_fn(pred, y)\n\n        optimizer.zero_grad()\n        loss.backward()\n        optimizer.step()\n\n        if batch % 500 == 0:\n            loss, current = loss.item(), batch * len(X)\n            print(f\"loss: {loss:>7f} [{current:>5d}/{size:>5d}]\")\n            \ndef validation(dataloader, model, loss_fn):\n    size = len(dataloader.dataset)\n    num_batches = len(dataloader)\n    model.eval()\n    test_loss, correct = 0, 0\n    with torch.no_grad():\n        for X, y in dataloader:\n            X, y = X.to(device), y.to(device)\n            pred = model(X)\n            test_loss += loss_fn(pred, y).item()\n            correct += (pred.argmax(1) == y).type(torch.float).sum().item()\n    test_loss /= num_batches\n    correct /= size\n    print(f\"Test Error: \\n Accuracy: {(100*correct):>0.1f}%, Avg loss: {test_loss:>8f}\\n\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#model = VanilaMyNet().to(device)\nmodel = resnet18_pretrained.to(device)\nprint(model)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"loss_fn = nn.CrossEntropyLoss()\noptimizer = torch.optim.Adam(model.parameters(), lr=1e-5)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"epochs = 5\nfor t in range(epochs):\n    print(f\"Epoch {t+1}\\n---------------------\")\n    train(train_dataloader, model, loss_fn, optimizer)\n    validation(validation_dataloader, model, loss_fn)\nprint(\"Done!\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Let's Predict!","metadata":{}},{"cell_type":"code","source":"test_dir = \"./test1\"\n# Make directorys which save data (For ImageFolder)\ntry:\n    if not os.path.exists(f\"{test_dir}/test\"):\n        os.makedirs(f\"{test_dir}/test\")\nexcept OSError:\n    print(\"ERROR : Creating Directory\")\n\nfor file_name in os.listdir(test_dir):\n    try:\n        shutil.move(f\"{test_dir}/{file_name}\", f\"{test_dir}/test/{file_name}\")\n    except:\n        continue\nos.listdir(test_dir)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test_dataset = torchvision.datasets.ImageFolder(root=test_dir, transform=data_transform)\ntest_dataloader = DataLoader(test_dataset, batch_size=1, shuffle=False)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"for value in train_dataloader:\n    data, _ = value\n    print(data.shape)\n    plt.imshow(torch.squeeze(data).permute(1, 2, 0))\n    break","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def test(dataloader, model, y_pred_output):\n    size = len(dataloader.dataset)\n    num_batches = len(dataloader)\n    model.eval()\n    with torch.no_grad():\n        for value in dataloader:\n            X, _ = value\n            X = X.to(device)\n            y_pred = model(X).cpu().data.numpy().argmax()\n            y_pred_output.append(y_pred)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"y_pred_output = []\ntest(test_dataloader, model, y_pred_output)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sample = pd.read_csv(\"../input/dogs-vs-cats/sampleSubmission.csv\")\ni = sample.id\n\n# dog = 1, cat = 0\ndef get_label(pred):\n    return 1 if idx_to_class[pred] == \"dog\" else 0\n\noutput = pd.DataFrame({'id' : i, 'label':list(map(get_label, y_pred_output))})\noutput.head(20)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"output.to_csv('my_submission.csv', index=False)","metadata":{"trusted":true},"execution_count":null,"outputs":[]}]}