{"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":"import numpy as np\nimport pandas as pd\nimport os\nimport cv2\nimport torch\nimport torch.nn as nn\nfrom torch.utils.data import Dataset, DataLoader\nfrom torchvision import transforms\nimport torchvision.models as models\nfrom tqdm.notebook import tqdm\n\nimport sys\nsys.path.append('../input/timm-pytorch-image-models/pytorch-image-models-master')\nimport timm\nimport albumentations as A\nfrom albumentations.pytorch.transforms import ToTensorV2\n\nTESTSAMPLEPATH = '../input/cassava-leaf-disease-classification/sample_submission.csv'\nCLASSESJSONPATH = '../input/cassava-leaf-disease-classification/label_num_to_disease_map.json'\nTESTDATAPATH = '../input/cassava-leaf-disease-classification/test_images'\nTRAINDATAPATH = '../input/cassava-leaf-disease-classification/train_images'\ncfg = {}\ncfg['image_size'] = 512 # (600, 800)\ncfg['device'] = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\ncfg['batch_size'] = 16\nclasses = {}\nnum_class = 0\n# get the class\nimport json\n\nwith open(CLASSESJSONPATH) as f:\n    classes = json.load(f)\n    num_class = len(classes)\n\ndef readImage(ID, path):\n    ID = ID.split('.')[0]\n    filepath = os.path.join(path, ID + '.jpg')\n    img = cv2.imread(filepath)\n    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n    return img","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2021-11-01T05:22:58.781224Z","iopub.execute_input":"2021-11-01T05:22:58.782109Z","iopub.status.idle":"2021-11-01T05:23:07.330111Z","shell.execute_reply.started":"2021-11-01T05:22:58.781986Z","shell.execute_reply":"2021-11-01T05:23:07.329278Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# dataset and data loader\nclass CLCdataset(Dataset):\n    def __init__(self, df,isTrain = True):\n        self.df = df\n        self.isTrain = isTrain\n        \n    def __len__(self):\n        return len(self.df)\n    \n    def __getitem__(self, idx):\n        ID = self.df['image_id'].iloc[idx].split('.')[0]\n        img = readImage(ID, TRAINDATAPATH if self.isTrain else TESTDATAPATH)\n        if self.isTrain:\n            img = train_transform(image = img)['image']\n        else:\n            img = test_transform(image = img)['image']\n        if self.isTrain:\n            label = self.df['label'].iloc[idx]\n            return ID, img, label\n        return ID, img\n    ","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2021-11-01T05:23:07.333327Z","iopub.execute_input":"2021-11-01T05:23:07.333948Z","iopub.status.idle":"2021-11-01T05:23:07.341627Z","shell.execute_reply.started":"2021-11-01T05:23:07.333918Z","shell.execute_reply":"2021-11-01T05:23:07.340695Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test_transform = A.Compose([\n    A.Resize(cfg['image_size'], cfg['image_size'],p=1),\n    A.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225], max_pixel_value=255.0, p=1.0),\n    ToTensorV2() # (H,W,C) to (C,H,W)\n])","metadata":{"execution":{"iopub.status.busy":"2021-11-01T05:23:07.343984Z","iopub.execute_input":"2021-11-01T05:23:07.344491Z","iopub.status.idle":"2021-11-01T05:23:07.354897Z","shell.execute_reply.started":"2021-11-01T05:23:07.344442Z","shell.execute_reply":"2021-11-01T05:23:07.354176Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"class CFCmodel(nn.Module):\n    def __init__(self, out_dim):\n        super(CFCmodel, self).__init__()\n        self.model = timm.create_model('efficientnet_b3', pretrained=False)\n        self.model.classifier = nn.Linear(in_features=1536,out_features=out_dim, bias=True)\n        self.model.eval()\n#         self.classifier = nn.Sequential(nn.Linear(1000, out_dim),\n#                                         nn.Softmax())\n#         self.classifier = nn.Sequential(nn.Linear(1000, out_dim))\n        # freeze attibute\n#         for p in self.features.parameters():\n#             p.requires_grad = False\n    def forward(self, input):\n        return self.model(input)","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2021-11-01T05:24:22.763417Z","iopub.execute_input":"2021-11-01T05:24:22.763676Z","iopub.status.idle":"2021-11-01T05:24:22.770623Z","shell.execute_reply.started":"2021-11-01T05:24:22.763645Z","shell.execute_reply":"2021-11-01T05:24:22.769827Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"\ndef test(model, dataloader):\n    model.eval()\n    pred_list = []\n    pred_id_list = []\n    with torch.no_grad():\n        with tqdm(dataloader,unit='batch',desc='Test') as tqdm_loader:\n            for idx, (ID, img) in enumerate(tqdm_loader):\n\n                img = img.to(device=cfg['device'])\n                pred = model(img).detach().cpu().argmax(dim=1)\n\n                pred_list.append(pred)\n                pred_id_list.append(ID)\n    pred_list = np.concatenate(pred_list,axis=0)\n    pred_id_list =  np.concatenate(pred_id_list,axis=0)\n    return pred_list, pred_id_list\n\ntest_df = pd.read_csv(TESTSAMPLEPATH)\ntest_dataset = CLCdataset(test_df,isTrain=False)\ntest_dataloader = DataLoader(test_dataset, batch_size=cfg['batch_size'], shuffle=False, num_workers=2)\n\n\nmodel = CFCmodel(num_class).to(cfg['device'])\n\nmodel.load_state_dict(torch.load('../input/k/moore0403/cfdc-competition/model0_6.pth'))\n\npred_list, pred_id_list = test(model, test_dataloader)\n\npred_id_list = [ID +'.jpg' for ID in pred_id_list]\n\nsubmit_df  = pd.DataFrame(columns=test_df.columns)\nsubmit_df['image_id'] = pred_id_list\nsubmit_df['label']=pred_list\nsubmit_df.to_csv(os.path.join('./', 'submission.csv'),index=False)\n# print(submit_df.head(10))","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2021-11-01T05:24:26.628091Z","iopub.execute_input":"2021-11-01T05:24:26.628630Z","iopub.status.idle":"2021-11-01T05:24:38.650860Z","shell.execute_reply.started":"2021-11-01T05:24:26.628590Z","shell.execute_reply":"2021-11-01T05:24:38.650112Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]}]}