{"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":"!pip install efficientnet_pytorch","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import numpy as np \nimport pandas as pd \nimport shutil\nimport os\nimport zipfile\nimport torch\nimport torch.nn as nn\nimport cv2\nimport matplotlib.pyplot as plt\nimport torchvision\nfrom torch.utils.data import Dataset, DataLoader\nfrom torch.utils.data.sampler import SubsetRandomSampler\nfrom torchvision import transforms\nimport torch.nn.functional as F\nimport copy\nimport tqdm\nimport time\nfrom PIL import Image\n\nimport albumentations\nfrom albumentations import pytorch as AT\n\nfrom efficientnet_pytorch import EfficientNet\n\n%matplotlib inline","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_dir = '../input/plant-pathology-2021-224x224/train_imgs'\ntest_dir = '../input/plant-pathology-2021-fgvc8/test_images'","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_data = pd.read_csv('../input/plant-pathology-2021-fgvc8/train.csv')\na = np.asarray(train_data['labels'].unique())","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"len(a)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"from sklearn.preprocessing import LabelEncoder","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"encoder = LabelEncoder()\ntrain_data['num_labels'] = encoder.fit_transform(train_data['labels'])","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_data","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_files = os.listdir(train_dir)\ntest_f = os.listdir(test_dir)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test_files = pd.DataFrame()\ntest_files['image'] = test_f\ntest_files","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_files = train_data\ntrain_files","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"class CustomDataset(Dataset):\n    def __init__(self, file_csv, dir, transform=None, mode = 'train'):\n        self.file_csv = file_csv\n        self.dir = dir\n        self.transform = transform\n        self.mode = mode\n            \n    def __len__(self):\n        return len(self.file_csv)\n    \n    #метод который позволяет нам индексировать датасет\n    def __getitem__(self, idx):\n        #считываем изображение\n        image = cv2.imread(os.path.join(self.dir, self.file_csv['image'][idx]))\n        image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n        \n        self.label = int(train_data['num_labels'][idx])\n        \n        #применяем аугментации\n        if self.transform:\n            augmented = self.transform(image=image)\n            image = augmented['image']\n        \n        if self.mode == 'train':\n            return image, float(self.label)\n        else:\n            return image, self.file_csv['image'][idx]\n","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"batch_size = 16\nnum_workers = 0\nimg_size = 224","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"data_transforms_train = albumentations.Compose([\n    albumentations.Resize(img_size, img_size),\n    AT.ToTensor()\n    ])\n\ndata_transforms_test = albumentations.Compose([\n    albumentations.Resize(img_size, img_size),\n    AT.ToTensor()\n    ])","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"trainset = CustomDataset(train_files, train_dir, transform = data_transforms_train)\ntestset = CustomDataset(test_files, test_dir, transform=data_transforms_test, mode = 'test')","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"valid_size = int(len(train_files) * 0.1)\ntrain_set, valid_set = torch.utils.data.random_split(trainset, \n                                    (len(train_files)-valid_size, valid_size))","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"len(train_set), len(valid_set)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"trainloader = torch.utils.data.DataLoader(train_set, pin_memory=True, \n                                        batch_size=batch_size, shuffle=True)\nvalidloader = torch.utils.data.DataLoader(valid_set, pin_memory=True, \n                                        batch_size=batch_size, shuffle=True)\n\ntestloader = torch.utils.data.DataLoader(testset, batch_size = batch_size,\n                                         num_workers = num_workers)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\ndevice","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model = EfficientNet.from_pretrained('efficientnet-b3')","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model._fc = nn.Linear(in_features = 1536, out_features = 12)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def train_model(model_conv, train_loader, valid_loader, criterion, optimizer,scheduler ,n_epochs):\n    model_conv.to(device)\n    valid_loss_min = np.Inf\n    patience = 5\n    # сколько эпох ждем до отключения\n    p = 0\n    # иначе останавливаем обучение\n    stop = False\n\n    # количество эпох\n    for epoch in range(1, n_epochs+1):\n        print(time.ctime(), 'Epoch:', epoch)\n\n        train_loss = []\n\n        for batch_i, (data, target) in enumerate(train_loader):\n\n            data, target = data.to(device), target.to(device)\n\n            optimizer.zero_grad()\n            output = model_conv(data)\n            loss = criterion(output, target.long())\n            train_loss.append(loss.item())\n            loss.backward()\n            optimizer.step()\n    # запускаем валидацию\n        model_conv.eval()\n        correct = 0\n        val_loss = []\n        for batch_i, (data, target) in enumerate(valid_loader):\n            data, target = data.to(device), target.to(device)\n            output = model_conv(data)\n            _, predicted = torch.max(output.data, 1)\n            correct += (predicted == target).sum().item()\n            loss = criterion(output, target.long())\n            val_loss.append(loss.item()) \n        \n        acc = correct / len(valid_set)\n\n        print(f'Epoch {epoch}, train loss: {np.mean(train_loss):.4f}, valid loss: {np.mean(val_loss):.4f}.')\n        print(f'Accuracy on valid set: {acc}')\n\n        valid_loss = np.mean(val_loss)\n        scheduler.step(valid_loss)\n        if valid_loss <= valid_loss_min:\n            print('Validation loss decreased ({:.6f} --> {:.6f}).  Saving model ...'.format(\n            valid_loss_min,\n            valid_loss))\n            torch.save(model_conv.state_dict(), 'model.pt')\n            valid_loss_min = valid_loss\n            p = 0\n\n        # проверяем как дела на валидации\n        if valid_loss > valid_loss_min:\n            p += 1\n            print(f'{p} epochs of increasing val loss')\n            if p > patience:\n                print('Stopping training')\n                stop = True\n                break        \n\n        if stop:\n            break\n    return model_conv, train_loss, val_loss","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"criterion = nn.CrossEntropyLoss()\noptimizer = torch.optim.Adam(model.parameters(), lr=0.00001)\nscheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer, factor=0.8, patience=2,)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# model_, train_loss, val_loss = train_model(model, trainloader, validloader, criterion, \n#                               optimizer,scheduler, n_epochs=15)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# pd.read_csv('../input/plant-pathology-2021-fgvc8/sample_submission.csv')","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model = EfficientNet.from_pretrained('efficientnet-b3')\nmodel._fc = nn.Linear(in_features = 1536, out_features = 12)\n\nmodel.state_dict(torch.load('./model.pt'))","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"from pathlib import Path\n\nmodel.to(device)\nmodel.eval()\ndf_preds = pd.DataFrame()\nfor path in Path('../input/plant-pathology-2021-fgvc8/test_images').iterdir():\n    img = cv2.imread(str(path))[:, ::-1]\n    img = data_transforms_test(image=img)['image'].cuda()\n    pred = model(img[None])\n    \n    df_preds = df_preds.append(\n        {'image': path.parts[-1], 'labels': encoder.inverse_transform(torch.argmax(pred.cpu(), dim=1))[0]},\n        ignore_index=True)\n    \ndf_preds","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df_preds.to_csv('./submission.csv', index = False)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]}]}