{"cells":[{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"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 in \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 \"../input/\" directory.\n# For example, running this (by clicking run or pressing Shift+Enter) will list the files in the input directory\n\nimport os\nprint(os.listdir(\"../input\"))\n\n# Any results you write to the current directory are saved as output.","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","trusted":true},"cell_type":"code","source":"def My_TTA(model, TTA=3, batch_size=128, threshold = 0.1):\n    \n    model.eval()\n    avg_predictions = {}\n    ans_dict = {}\n    models_num = len(os.listdir(\"../input/your_models_folder\"))\n    \n    for time in range(TTA):\n        \n        test_transformed_dataset = iMetDataset(csv_file='sample_submission.csv', \n                                      label_file=\"labels.csv\", \n                                      img_path=\"test/\", \n                                      root_dir='../input/imet-2019-fgvc6/',\n                                      transform=transforms.Compose([\n                                          #\n                                          # some data augumentations here\n                                          #\n                                          transforms.ToTensor(),\n                                          transforms.Normalize(\n                                              [0.485, 0.456, 0.406], \n                                              [0.229, 0.224, 0.225])\n                                      ]))\n\n        test_loader = DataLoader(\n        test_transformed_dataset,\n        batch_size=batch_size,\n        shuffle=False,\n        num_workers=8)\n\n        with torch.no_grad():\n            \n            for i in range(models_num):\n\n                model.load_state_dict(torch.load(\"../input/your_models_folder/your_model\" + str(i)+ \".pth\"))\n   \n                for batch_idx, sample in enumerate(test_loader):\n     \n                    image = sample[\"image\"].to(device, dtype=torch.float)\n                    img_ids = sample[\"img_id\"]\n                    predictions = model(image).cpu().numpy()\n                    \n                    for row, img_id in enumerate(img_ids):\n                        if time == 0 and i == 0:\n                            avg_predictions[img_id] = predictions[row]/(TTA*models_num)\n                        else:\n                            avg_predictions[img_id] += predictions[row]/(TTA*models_num)\n\n                        if time == TTA - 1 and i == models_num -1:\n                            all_class = np.nonzero(avg_predictions[img_id] > threshold)[0].tolist()\n                            all_class = [str(x) for x in all_class]\n                            ans_dict[img_id] = \" \".join(all_class)\n    \n    return ans_dict","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"name":"python","version":"3.6.4","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat":4,"nbformat_minor":1}