{"cells":[{"metadata":{"trusted":true},"cell_type":"code","source":"%matplotlib inline\nimport os\nimport pandas as pd\nfrom glob import glob\nimport numpy as np\nfrom fastai import *\nfrom fastai.vision import *\nimport torch \nimport matplotlib.pyplot as plt","execution_count":null,"outputs":[]},{"metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true},"cell_type":"code","source":"path = Path('../input/aptos2019-blindness-detection')\npath_train = path/'train_images'\npath_test = path/'test_images'\npath, path_train, path_test","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"labels = pd.read_csv(path/'train.csv')\nlabels.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"img = open_image(path_train/'000c1434d8d7.png')\nimg.show(figsize = (7,7))\nprint(img.shape)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"labels['diagnosis'].value_counts().plot(kind = 'bar', title='Distribution of diagnosis categories')\nplt.show()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"tfms = get_transforms(\n    do_flip=True,\n    flip_vert=True,\n    max_warp=0.1,\n    max_rotate=66.,\n    max_zoom=1.1,\n    max_lighting=0.1,\n    p_lighting=0.5\n)\naptos19_stats = ([0.42, 0.22, 0.075], [0.27, 0.15, 0.081])","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"test_labels = pd.read_csv(path/'sample_submission.csv')\ntest = ImageList.from_df(test_labels, path = path_test, suffix = '.png')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"src = (ImageList.from_df(labels, path = path_train, suffix = '.png')\n       .split_by_rand_pct(seed = 42)\n       .label_from_df(cols = 'diagnosis')\n       .add_test(test))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"data = (\n    src.transform(\n        tfms,\n        size = 128, \n        resize_method=ResizeMethod.SQUISH,\n        padding_mode='zeros'\n    )\n    .databunch(bs=32)\n    .normalize(aptos19_stats))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"data.show_batch(3, figsize = (7,7))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"print(data.classes)\nprint(len(data.train_ds))\nprint(len(data.valid_ds))\nprint(len(data.test_ds))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"!mkdir models\n!mkdir -p /tmp/.cache/torch/checkpoints\n!cp ../input/resnet34/resnet34.pth /tmp/.cache/torch/checkpoints/resnet34-333f7ec4.pth","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"kappa = KappaScore()\nkappa.weights = \"quadratic\"","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"learn = cnn_learner(\n    data, \n    models.resnet34, \n    metrics = [accuracy, kappa], \n    model_dir = Path('../kaggle/working'),\n    path = Path(\".\")\n)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"learn.fit_one_cycle(15)\nlearn.save('resnet34')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"learn.load('resnet34')\nlearn.unfreeze()\nlearn.lr_find()\nlearn.recorder.plot()\nlearn.fit_one_cycle(20, slice(1e-6,5e-4))\nlearn.save('resnet34-1')","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"## Double the size of images"},{"metadata":{"trusted":true},"cell_type":"code","source":"# learn.load('resnet152-2')\n# data = (\n#     src.transform(\n#         tfms,\n#         size = 1024, \n#         resize_method=ResizeMethod.SQUISH,\n#         padding_mode='zeros'\n#     )\n#     .databunch(bs=4)\n#     .normalize(aptos19_stats))\n# learn.data = data\n# learn.freeze()\n# learn.lr_find()\n# learn.recorder.plot()\n# learn.fit_one_cycle(4, 2e-4)\n# learn.save('resnet152-3')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# learn.load('resnet152-3')\n# learn.unfreeze()\n# learn.lr_find()\n# learn.recorder.plot()\n# learn.fit_one_cycle(6, 2e-5)\n# learn.save('resnet152-4')\n","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"## Preparing Submission"},{"metadata":{"trusted":true},"cell_type":"code","source":"learn.load('resnet34-1')\npreds, _ = learn.get_preds(ds_type=DatasetType.Test)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"submission = pd.read_csv(path/'sample_submission.csv')\nsubmission.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"preds = np.array(preds.argmax(1)).astype(int).tolist()\npreds[:5]","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"submission['diagnosis'] = preds\nsubmission.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"import warnings\nwarnings.simplefilter(action='ignore', category=FutureWarning)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"submission.to_csv('submission.csv', index = False)","execution_count":null,"outputs":[]}],"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":4,"nbformat_minor":1}