my-kaggle-cassava-0-data-kfold.ipynb

{"cells":[{"metadata":{"execution":{"iopub.execute_input":"2021-01-05T18:28:01.371112Z","iopub.status.busy":"2021-01-05T18:28:01.370125Z","iopub.status.idle":"2021-01-05T18:28:01.372669Z","shell.execute_reply":"2021-01-05T18:28:01.373214Z"},"papermill":{"duration":0.037425,"end_time":"2021-01-05T18:28:01.373369","exception":false,"start_time":"2021-01-05T18:28:01.335944","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"#meta 1/17/2020 Kaggle Cassava Data\n#fast.ai 2020 study group Kaggle Competition\n#src https://www.kaggle.com/c/cassava-leaf-disease-classification\n#prev competition src https://www.kaggle.com/c/cassava-disease\n#references https://arxiv.org/pdf/1908.02900.pdf\n\n#input: kaggle files - images, csv with 5 cats\n#output: data artifacts for next step 2-model\n#pwd: /kaggle/working\n#based on my-kaggle-cassava-1-model.ipynb\n\n#history\n#1/4/2021 DATA - IMBALANCED DS -> STRATIFIED VALIDATION SET\n#      dataprep delta: Stratified data split\n#      src https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.StratifiedShuffleSplit.html\n#      dataprep delta: add df_train['is_valid'] column needed for ColSplitter()\n#      src https://docs.fast.ai/tutorial.datablock.html\n\n#1/17/2021 DATA MODULE, ORIGINAL TRAIN DS AND OVERSAMPLED TRAIN DS\n#      With Stratified data split\n#      Imbalanced dataset -> oversample\n#      Export train datasets. $param\n\n#2/7/2020 DATA MODULE, ORIGINAL TRAIN DS AND OVERSAMPLED TRAIN DS\n#      With Stratified data split\n#      Imbalanced dataset -> manual oversample\n#      2 versions: full and mini for rapid iteration (10%) $param\n\n#here 2/7/2020 DATA MODULE, K-FOLD STRATIFIED SPLIT\n#      Original full ds and oversampled ds, with Stratified data split\n#      Imbalanced dataset -> manual oversample\n#      2 versions: full and mini for rapid iteration (10%) $param\n#here 3/3/2020 FINALIZED K-FOLD VERSION","execution_count":16,"outputs":[]},{"metadata":{"_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","execution":{"iopub.execute_input":"2021-01-05T18:28:01.427494Z","iopub.status.busy":"2021-01-05T18:28:01.426674Z","iopub.status.idle":"2021-01-05T18:28:04.438988Z","shell.execute_reply":"2021-01-05T18:28:04.437608Z"},"papermill":{"duration":3.041862,"end_time":"2021-01-05T18:28:04.439159","exception":false,"start_time":"2021-01-05T18:28:01.397297","status":"completed"},"tags":[],"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\nimport os\nimport time #to track performance time\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n\n#stratified split\nfrom sklearn.model_selection import StratifiedShuffleSplit\nfrom sklearn.model_selection import StratifiedKFold\n\n#modeling\nfrom fastai.vision.all import *\n\n\n# Input data files are available in the read-only \"../input/\" directory\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","execution_count":17,"outputs":[]},{"metadata":{"execution":{"iopub.execute_input":"2021-01-05T18:28:04.499792Z","iopub.status.busy":"2021-01-05T18:28:04.498955Z","iopub.status.idle":"2021-01-05T18:28:04.504311Z","shell.execute_reply":"2021-01-05T18:28:04.501715Z"},"papermill":{"duration":0.036705,"end_time":"2021-01-05T18:28:04.504414","exception":false,"start_time":"2021-01-05T18:28:04.467709","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"#--constants and variables\nSEED = 24\nMINI_FLG = 0 #$param\nMINI_DS_SIZE = 0.1\nVALID_DS_SIZE = 0.2\n\npath = Path('../input/cassava-leaf-disease-classification')\n\n#--functions\n#define label lookup \ndef lookup_label(image_id): return int(df_train[df_train['image_id']== image_id]['label'])\n\n#test function\n#lookup_label('100042118.jpg')","execution_count":18,"outputs":[]},{"metadata":{"papermill":{"duration":0.033823,"end_time":"2021-01-05T18:28:04.564011","exception":false,"start_time":"2021-01-05T18:28:04.530188","status":"completed"},"tags":[]},"cell_type":"markdown","source":"# Cassava Data\n21397 train images and 1 test image"},{"metadata":{"papermill":{"duration":0.026266,"end_time":"2021-01-05T18:28:04.616953","exception":false,"start_time":"2021-01-05T18:28:04.590687","status":"completed"},"tags":[]},"cell_type":"markdown","source":"## 0. Load Data\n- input images\n- output labels\n- train dataset"},{"metadata":{"execution":{"iopub.execute_input":"2021-01-05T18:28:04.673717Z","iopub.status.busy":"2021-01-05T18:28:04.673143Z","iopub.status.idle":"2021-01-05T18:28:04.681476Z","shell.execute_reply":"2021-01-05T18:28:04.682083Z"},"papermill":{"duration":0.0385,"end_time":"2021-01-05T18:28:04.682205","exception":false,"start_time":"2021-01-05T18:28:04.643705","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"path.ls()","execution_count":19,"outputs":[{"output_type":"execute_result","execution_count":19,"data":{"text/plain":"(#7) [Path('../input/cassava-leaf-disease-classification/train_tfrecords'),Path('../input/cassava-leaf-disease-classification/sample_submission.csv'),Path('../input/cassava-leaf-disease-classification/test_tfrecords'),Path('../input/cassava-leaf-disease-classification/label_num_to_disease_map.json'),Path('../input/cassava-leaf-disease-classification/train_images'),Path('../input/cassava-leaf-disease-classification/train.csv'),Path('../input/cassava-leaf-disease-classification/test_images')]"},"metadata":{}}]},{"metadata":{"execution":{"iopub.execute_input":"2021-01-05T18:28:04.792222Z","iopub.status.busy":"2021-01-05T18:28:04.791563Z","iopub.status.idle":"2021-01-05T18:28:05.160966Z","shell.execute_reply":"2021-01-05T18:28:05.160049Z"},"papermill":{"duration":0.399659,"end_time":"2021-01-05T18:28:05.161086","exception":false,"start_time":"2021-01-05T18:28:04.761427","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"(path/'train_images').ls(), (path/'test_images').ls()","execution_count":20,"outputs":[{"output_type":"execute_result","execution_count":20,"data":{"text/plain":"((#21397) [Path('../input/cassava-leaf-disease-classification/train_images/1235188286.jpg'),Path('../input/cassava-leaf-disease-classification/train_images/1215607589.jpg'),Path('../input/cassava-leaf-disease-classification/train_images/478554372.jpg'),Path('../input/cassava-leaf-disease-classification/train_images/2763304605.jpg'),Path('../input/cassava-leaf-disease-classification/train_images/2826122413.jpg'),Path('../input/cassava-leaf-disease-classification/train_images/111117998.jpg'),Path('../input/cassava-leaf-disease-classification/train_images/231268038.jpg'),Path('../input/cassava-leaf-disease-classification/train_images/4201965605.jpg'),Path('../input/cassava-leaf-disease-classification/train_images/3224710052.jpg'),Path('../input/cassava-leaf-disease-classification/train_images/1290729293.jpg')...],\n (#1) [Path('../input/cassava-leaf-disease-classification/test_images/2216849948.jpg')])"},"metadata":{}}]},{"metadata":{"execution":{"iopub.execute_input":"2021-01-05T18:28:05.218887Z","iopub.status.busy":"2021-01-05T18:28:05.218092Z","iopub.status.idle":"2021-01-05T18:28:23.057324Z","shell.execute_reply":"2021-01-05T18:28:23.056516Z"},"papermill":{"duration":17.869379,"end_time":"2021-01-05T18:28:23.057489","exception":false,"start_time":"2021-01-05T18:28:05.18811","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"#load train/test images\n#train_img = get_image_files(path/'train_images')\n#test_img = get_image_files(path/'test_images')\n\n#print(\"Number of files: \", (len(train_img), len(test_img)))","execution_count":21,"outputs":[]},{"metadata":{"execution":{"iopub.execute_input":"2021-01-05T18:28:23.171079Z","iopub.status.busy":"2021-01-05T18:28:23.170342Z","iopub.status.idle":"2021-01-05T18:28:23.179682Z","shell.execute_reply":"2021-01-05T18:28:23.179204Z"},"papermill":{"duration":0.042162,"end_time":"2021-01-05T18:28:23.179778","exception":false,"start_time":"2021-01-05T18:28:23.137616","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"#labels\nwith open((path/'label_num_to_disease_map.json')) as json_file:\n    target_map = json.load(json_file)\n    \nprint(target_map)","execution_count":22,"outputs":[{"output_type":"stream","text":"{'0': 'Cassava Bacterial Blight (CBB)', '1': 'Cassava Brown Streak Disease (CBSD)', '2': 'Cassava Green Mottle (CGM)', '3': 'Cassava Mosaic Disease (CMD)', '4': 'Healthy'}\n","name":"stdout"}]},{"metadata":{"execution":{"iopub.execute_input":"2021-01-05T18:28:23.298603Z","iopub.status.busy":"2021-01-05T18:28:23.297973Z","iopub.status.idle":"2021-01-05T18:28:23.339766Z","shell.execute_reply":"2021-01-05T18:28:23.34028Z"},"papermill":{"duration":0.075262,"end_time":"2021-01-05T18:28:23.340393","exception":false,"start_time":"2021-01-05T18:28:23.265131","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"#train Dataset\ndf_train = pd.read_csv(path/'train.csv')\n\ndf_train['k_fold'] = None\n\nprint('train data\\n', df_train.shape)\ndisplay(df_train.head())\n","execution_count":23,"outputs":[{"output_type":"stream","text":"train data\n (21397, 3)\n","name":"stdout"},{"output_type":"display_data","data":{"text/plain":"         image_id  label k_fold\n0  1000015157.jpg      0   None\n1  1000201771.jpg      3   None\n2   100042118.jpg      1   None\n3  1000723321.jpg      1   None\n4  1000812911.jpg      3   None","text/html":"<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>image_id</th>\n      <th>label</th>\n      <th>k_fold</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>1000015157.jpg</td>\n      <td>0</td>\n      <td>None</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>1000201771.jpg</td>\n      <td>3</td>\n      <td>None</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>100042118.jpg</td>\n      <td>1</td>\n      <td>None</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>1000723321.jpg</td>\n      <td>1</td>\n      <td>None</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>1000812911.jpg</td>\n      <td>3</td>\n      <td>None</td>\n    </tr>\n  </tbody>\n</table>\n</div>"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"## 1. Prep Data\n### 1.1 Full or Mini dataset"},{"metadata":{"execution":{"iopub.execute_input":"2021-01-05T18:28:23.405479Z","iopub.status.busy":"2021-01-05T18:28:23.404695Z","iopub.status.idle":"2021-01-05T18:28:23.408136Z","shell.execute_reply":"2021-01-05T18:28:23.407584Z"},"papermill":{"duration":0.039099,"end_time":"2021-01-05T18:28:23.408237","exception":false,"start_time":"2021-01-05T18:28:23.369138","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"#train mini subsample for rapid iteration\nif MINI_FLG:\n    #stratified subsample\n    X_mini = df_train['image_id']\n    y_mini = df_train['label']\n    sss_mini = StratifiedShuffleSplit(n_splits=1, test_size=MINI_DS_SIZE, random_state=SEED)\n    \n    for train_idx_mini, valid_idx_mini in sss_mini.split(X_mini, y_mini):\n        print(\"TRAIN:\", train_idx_mini, \"VALID:\", valid_idx_mini)\n        y_train_mini, y_valid_mini = y_mini[train_idx_mini], y_mini[valid_idx_mini]\n\n    y_train_mini.value_counts().sort_index()\n    y_train_mini.value_counts().sort_index().plot(kind='bar');\n    ","execution_count":24,"outputs":[{"output_type":"stream","text":"TRAIN: [20562  3655 13241 ...  2664  2076  4573] VALID: [14727 12496 11627 ...  9559 12465  1439]\n","name":"stdout"},{"output_type":"display_data","data":{"text/plain":"<Figure size 432x288 with 1 Axes>","image/png":"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\n"},"metadata":{"needs_background":"light"}}]},{"metadata":{"execution":{"iopub.execute_input":"2021-01-05T18:28:23.471562Z","iopub.status.busy":"2021-01-05T18:28:23.470714Z","iopub.status.idle":"2021-01-05T18:28:23.473699Z","shell.execute_reply":"2021-01-05T18:28:23.473085Z"},"papermill":{"duration":0.03719,"end_time":"2021-01-05T18:28:23.473788","exception":false,"start_time":"2021-01-05T18:28:23.436598","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"#train subsample\nif MINI_FLG:\n    #update df_train with train index \n    df_train = df_train.loc[df_train.index.isin(valid_idx_mini)].reset_index()\n    print(df_train.shape)\n    print(df_train.head())\n    ","execution_count":25,"outputs":[{"output_type":"stream","text":"(2140, 4)\n   index        image_id  label k_fold\n0     19  1003987001.jpg      3   None\n1     23  1004672608.jpg      3   None\n2     24   100472565.jpg      2   None\n3     27  1005138819.jpg      3   None\n4     42  1008244905.jpg      1   None\n","name":"stdout"}]},{"metadata":{"execution":{"iopub.execute_input":"2021-01-05T18:28:23.536489Z","iopub.status.busy":"2021-01-05T18:28:23.535777Z","iopub.status.idle":"2021-01-05T18:28:23.727939Z","shell.execute_reply":"2021-01-05T18:28:23.727298Z"},"papermill":{"duration":0.2259,"end_time":"2021-01-05T18:28:23.728071","exception":false,"start_time":"2021-01-05T18:28:23.502171","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"print(df_train['label'].value_counts().sort_index())\ndf_train['label'].value_counts().sort_index().plot(kind='bar');","execution_count":26,"outputs":[{"output_type":"stream","text":"0     109\n1     219\n2     238\n3    1316\n4     258\nName: label, dtype: int64\n","name":"stdout"},{"output_type":"display_data","data":{"text/plain":"<Figure size 432x288 with 1 Axes>","image/png":"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\n"},"metadata":{"needs_background":"light"}}]},{"metadata":{"papermill":{"duration":0.041134,"end_time":"2021-01-05T18:28:24.256934","exception":false,"start_time":"2021-01-05T18:28:24.2158","status":"completed"},"tags":[]},"cell_type":"markdown","source":"### 1.2 K-fold Stratified Split\nStratified K-Folds cross-validator.\n\nProvides train/test indices to split data in train/test sets.  This cross-validation object is a variation of KFold that returns stratified folds. The folds are made by preserving the percentage of samples for each class."},{"metadata":{"execution":{"iopub.execute_input":"2021-01-05T18:28:24.347437Z","iopub.status.busy":"2021-01-05T18:28:24.346586Z","iopub.status.idle":"2021-01-05T18:28:24.350838Z","shell.execute_reply":"2021-01-05T18:28:24.351376Z"},"papermill":{"duration":0.052339,"end_time":"2021-01-05T18:28:24.351502","exception":false,"start_time":"2021-01-05T18:28:24.299163","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"if MINI_FLG:\n    X = df_train[['index', 'image_id', 'k_fold']]\nelse:\n    X = df_train[['image_id', 'k_fold']]\ny = df_train['label']\n\nskf = StratifiedKFold(n_splits=5, shuffle=True, random_state=SEED)\nskf.get_n_splits(X, y)\nprint(skf)\n","execution_count":27,"outputs":[{"output_type":"stream","text":"StratifiedKFold(n_splits=5, random_state=24, shuffle=True)\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"k_fold = 0\n\nfor train_idx, valid_idx in skf.split(X, y):\n    #print(\"TRAIN:\", train_idx, \"\\nVALID:\", valid_idx)\n     \n    #update df_train with k-fold i\n    df_train.loc[(df_train.index.isin(valid_idx)), 'k_fold'] = k_fold\n    \n    k_fold+=1\n\n#preview df with k-fold index\ndf_train[::200]","execution_count":28,"outputs":[{"output_type":"execute_result","execution_count":28,"data":{"text/plain":"      index        image_id  label k_fold\n0        19  1003987001.jpg      3      2\n200    1963  1345734406.jpg      2      0\n400    3899  1694968228.jpg      3      4\n600    5705  2015518840.jpg      3      0\n800    7819  2391557874.jpg      4      2\n1000   9852  2752015224.jpg      3      0\n1200  11960  3141282286.jpg      2      3\n1400  14189   353518711.jpg      2      1\n1600  16010  3880254716.jpg      0      1\n1800  17921  4225506393.jpg      3      1\n2000  20041   736522598.jpg      2      1","text/html":"<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>index</th>\n      <th>image_id</th>\n      <th>label</th>\n      <th>k_fold</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>19</td>\n      <td>1003987001.jpg</td>\n      <td>3</td>\n      <td>2</td>\n    </tr>\n    <tr>\n      <th>200</th>\n      <td>1963</td>\n      <td>1345734406.jpg</td>\n      <td>2</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>400</th>\n      <td>3899</td>\n      <td>1694968228.jpg</td>\n      <td>3</td>\n      <td>4</td>\n    </tr>\n    <tr>\n      <th>600</th>\n      <td>5705</td>\n      <td>2015518840.jpg</td>\n      <td>3</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>800</th>\n      <td>7819</td>\n      <td>2391557874.jpg</td>\n      <td>4</td>\n      <td>2</td>\n    </tr>\n    <tr>\n      <th>1000</th>\n      <td>9852</td>\n      <td>2752015224.jpg</td>\n      <td>3</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>1200</th>\n      <td>11960</td>\n      <td>3141282286.jpg</td>\n      <td>2</td>\n      <td>3</td>\n    </tr>\n    <tr>\n      <th>1400</th>\n      <td>14189</td>\n      <td>353518711.jpg</td>\n      <td>2</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>1600</th>\n      <td>16010</td>\n      <td>3880254716.jpg</td>\n      <td>0</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>1800</th>\n      <td>17921</td>\n      <td>4225506393.jpg</td>\n      <td>3</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>2000</th>\n      <td>20041</td>\n      <td>736522598.jpg</td>\n      <td>2</td>\n      <td>1</td>\n    </tr>\n  </tbody>\n</table>\n</div>"},"metadata":{}}]},{"metadata":{"execution":{"iopub.execute_input":"2021-01-05T18:28:24.688443Z","iopub.status.busy":"2021-01-05T18:28:24.687372Z","iopub.status.idle":"2021-01-05T18:28:24.822704Z","shell.execute_reply":"2021-01-05T18:28:24.823271Z"},"papermill":{"duration":0.191146,"end_time":"2021-01-05T18:28:24.823422","exception":false,"start_time":"2021-01-05T18:28:24.632276","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"#k-fold counts\ndf_train.groupby('k_fold').count()","execution_count":29,"outputs":[{"output_type":"execute_result","execution_count":29,"data":{"text/plain":"        index  image_id  label\nk_fold                        \n0         428       428    428\n1         428       428    428\n2         428       428    428\n3         428       428    428\n4         428       428    428","text/html":"<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>index</th>\n      <th>image_id</th>\n      <th>label</th>\n    </tr>\n    <tr>\n      <th>k_fold</th>\n      <th></th>\n      <th></th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>428</td>\n      <td>428</td>\n      <td>428</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>428</td>\n      <td>428</td>\n      <td>428</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>428</td>\n      <td>428</td>\n      <td>428</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>428</td>\n      <td>428</td>\n      <td>428</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>428</td>\n      <td>428</td>\n      <td>428</td>\n    </tr>\n  </tbody>\n</table>\n</div>"},"metadata":{}}]},{"metadata":{"trusted":true},"cell_type":"code","source":"#export data\nif MINI_FLG:\n    torch.save(df_train, 'df_train_mini_kfold.pkl')\nelse:\n    torch.save(df_train, 'df_train_kfold.pkl')","execution_count":30,"outputs":[]}],"metadata":{"kernelspec":{"name":"python3","display_name":"Python 3","language":"python"},"language_info":{"name":"python","version":"3.7.9","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat":4,"nbformat_minor":4}