{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.7.12","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"import os, sys, math, json\nimport numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\nimport skimage.io\n\nfrom skimage.transform import resize\nfrom imgaug import augmenters as iaa\nfrom tqdm import tqdm\nimport PIL\nfrom PIL import Image, ImageOps\nimport cv2\nfrom sklearn.utils import class_weight, shuffle\n\n%matplotlib inline\n\nimport scipy\nimport tensorflow as tf\n\nfrom tensorflow.keras.losses import binary_crossentropy\nfrom tensorflow.keras.applications.resnet50 import preprocess_input\nimport tensorflow.keras.backend as K\nfrom tensorflow.keras.utils import Sequence\nfrom tensorflow.keras.utils import to_categorical\n\nfrom tensorflow.keras import layers\nfrom tensorflow.keras.applications import DenseNet121\nfrom tensorflow.keras.callbacks import Callback, ModelCheckpoint\nfrom tensorflow.keras.preprocessing.image import ImageDataGenerator\nfrom tensorflow.keras.models import Sequential\nfrom tensorflow.keras.optimizers import Adam\n\nfrom sklearn.metrics import f1_score, fbeta_score\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.metrics import cohen_kappa_score, accuracy_score\n\n\nWORKERS = 2\nCHANNEL = 3\n\nimport warnings\nwarnings.filterwarnings(\"ignore\")\n\nnp.random.seed(42)\ntf.random.set_seed(42)\n\nIMG_SIZE = 512\nNUM_CLASSES = 5\nSEED = 42\nTRAIN_NUM = 1000 # use 1000 when you just want to explore new idea, use -1 for full train\n\n#### Starting with just 2019 data","metadata":{"execution":{"iopub.status.busy":"2023-08-03T13:07:43.906208Z","iopub.execute_input":"2023-08-03T13:07:43.907257Z","iopub.status.idle":"2023-08-03T13:07:43.932293Z","shell.execute_reply.started":"2023-08-03T13:07:43.907203Z","shell.execute_reply":"2023-08-03T13:07:43.930792Z"},"trusted":true},"execution_count":2,"outputs":[]},{"cell_type":"code","source":"df_train = pd.read_csv('../input/diabetic-retinopathy-resized/trainLabels.csv')\ndf_test = pd.read_csv('../input/aptos2019-blindness-detection/test.csv')\n\nprint(df_train.shape)\nprint(df_test.shape)\nprint(df_train.head())\ndf_test","metadata":{"execution":{"iopub.status.busy":"2023-08-03T13:07:50.088834Z","iopub.execute_input":"2023-08-03T13:07:50.089281Z","iopub.status.idle":"2023-08-03T13:07:50.18294Z","shell.execute_reply.started":"2023-08-03T13:07:50.089243Z","shell.execute_reply":"2023-08-03T13:07:50.181521Z"},"trusted":true},"execution_count":3,"outputs":[{"name":"stdout","text":"(35126, 2)\n(1928, 1)\n image level\n0 10_left 0\n1 10_right 0\n2 13_left 0\n3 13_right 0\n4 15_left 1\n","output_type":"stream"},{"execution_count":3,"output_type":"execute_result","data":{"text/plain":" id_code\n0 0005cfc8afb6\n1 003f0afdcd15\n2 006efc72b638\n3 00836aaacf06\n4 009245722fa4\n... ...\n1923 ff2fd94448de\n1924 ff4c945d9b17\n1925 ff64897ac0d8\n1926 ffa73465b705\n1927 ffdc2152d455\n\n[1928 rows x 1 columns]","text/html":"\n\n
\n \n \n | \n id_code | \n
\n \n \n \n | 0 | \n 0005cfc8afb6 | \n
\n \n | 1 | \n 003f0afdcd15 | \n
\n \n | 2 | \n 006efc72b638 | \n
\n \n | 3 | \n 00836aaacf06 | \n
\n \n | 4 | \n 009245722fa4 | \n
\n \n | ... | \n ... | \n
\n \n | 1923 | \n ff2fd94448de | \n
\n \n | 1924 | \n ff4c945d9b17 | \n
\n \n | 1925 | \n ff64897ac0d8 | \n
\n \n | 1926 | \n ffa73465b705 | \n
\n \n | 1927 | \n ffdc2152d455 | \n
\n \n
\n
1928 rows × 1 columns
\n
","image/png":"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\n"},"metadata":{}}]},{"cell_type":"code","source":"class_weights = df_train['level'].value_counts()\ndfs = [df_train[df_train['level'] == i].sample(5*(2 if i < 3 else 1),replace=True) for i in range(5)]\nresampled = pd.concat(dfs, axis = 0).reset_index(drop=True)\nresampled","metadata":{"execution":{"iopub.status.busy":"2023-08-03T13:08:00.289662Z","iopub.execute_input":"2023-08-03T13:08:00.290121Z","iopub.status.idle":"2023-08-03T13:08:00.324207Z","shell.execute_reply.started":"2023-08-03T13:08:00.290083Z","shell.execute_reply":"2023-08-03T13:08:00.323096Z"},"trusted":true},"execution_count":4,"outputs":[{"execution_count":4,"output_type":"execute_result","data":{"text/plain":" image level\n0 40740_right 0\n1 27179_left 0\n2 1422_right 0\n3 9277_right 0\n4 37301_left 0\n5 20546_right 0\n6 19356_left 0\n7 38205_right 0\n8 10814_right 0\n9 29059_left 0\n10 21833_left 1\n11 5944_right 1\n12 26433_right 1\n13 38925_left 1\n14 2531_left 1\n15 30402_right 1\n16 13490_right 1\n17 43450_left 1\n18 27077_right 1\n19 44263_left 1\n20 42237_right 2\n21 9690_right 2\n22 37846_left 2\n23 27779_right 2\n24 34173_left 2\n25 40318_left 2\n26 23531_right 2\n27 4043_left 2\n28 8716_right 2\n29 20799_left 2\n30 34547_left 3\n31 23957_left 3\n32 35285_left 3\n33 38662_right 3\n34 10489_right 3\n35 42815_left 4\n36 35766_right 4\n37 35861_right 4\n38 15870_right 4\n39 32630_left 4","text/html":"\n\n
\n \n \n | \n image | \n level | \n
\n \n \n \n | 0 | \n 40740_right | \n 0 | \n
\n \n | 1 | \n 27179_left | \n 0 | \n
\n \n | 2 | \n 1422_right | \n 0 | \n
\n \n | 3 | \n 9277_right | \n 0 | \n
\n \n | 4 | \n 37301_left | \n 0 | \n
\n \n | 5 | \n 20546_right | \n 0 | \n
\n \n | 6 | \n 19356_left | \n 0 | \n
\n \n | 7 | \n 38205_right | \n 0 | \n
\n \n | 8 | \n 10814_right | \n 0 | \n
\n \n | 9 | \n 29059_left | \n 0 | \n
\n \n | 10 | \n 21833_left | \n 1 | \n
\n \n | 11 | \n 5944_right | \n 1 | \n
\n \n | 12 | \n 26433_right | \n 1 | \n
\n \n | 13 | \n 38925_left | \n 1 | \n
\n \n | 14 | \n 2531_left | \n 1 | \n
\n \n | 15 | \n 30402_right | \n 1 | \n
\n \n | 16 | \n 13490_right | \n 1 | \n
\n \n | 17 | \n 43450_left | \n 1 | \n
\n \n | 18 | \n 27077_right | \n 1 | \n
\n \n | 19 | \n 44263_left | \n 1 | \n
\n \n | 20 | \n 42237_right | \n 2 | \n
\n \n | 21 | \n 9690_right | \n 2 | \n
\n \n | 22 | \n 37846_left | \n 2 | \n
\n \n | 23 | \n 27779_right | \n 2 | \n
\n \n | 24 | \n 34173_left | \n 2 | \n
\n \n | 25 | \n 40318_left | \n 2 | \n
\n \n | 26 | \n 23531_right | \n 2 | \n
\n \n | 27 | \n 4043_left | \n 2 | \n
\n \n | 28 | \n 8716_right | \n 2 | \n
\n \n | 29 | \n 20799_left | \n 2 | \n
\n \n | 30 | \n 34547_left | \n 3 | \n
\n \n | 31 | \n 23957_left | \n 3 | \n
\n \n | 32 | \n 35285_left | \n 3 | \n
\n \n | 33 | \n 38662_right | \n 3 | \n
\n \n | 34 | \n 10489_right | \n 3 | \n
\n \n | 35 | \n 42815_left | \n 4 | \n
\n \n | 36 | \n 35766_right | \n 4 | \n
\n \n | 37 | \n 35861_right | \n 4 | \n
\n \n | 38 | \n 15870_right | \n 4 | \n
\n \n | 39 | \n 32630_left | \n 4 | \n
\n \n
\n