nc-severstal08-segment-unetdataaugloss-8.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:39.799274Z",
     "iopub.status.busy": "2021-03-21T04:01:39.798605Z",
     "iopub.status.idle": "2021-03-21T04:01:45.770598Z",
     "shell.execute_reply": "2021-03-21T04:01:45.769495Z"
    },
    "papermill": {
     "duration": 5.99558,
     "end_time": "2021-03-21T04:01:45.770818",
     "exception": false,
     "start_time": "2021-03-21T04:01:39.775238",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import gc\n",
    "\n",
    "import cv2\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import backend as K\n",
    "from tensorflow.keras import layers\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from tensorflow.keras.models import Model, load_model\n",
    "from tensorflow.keras.layers import Input\n",
    "from tensorflow.keras.layers import Conv2D, Conv2DTranspose\n",
    "from tensorflow.keras.layers import MaxPooling2D\n",
    "from tensorflow.keras.layers import concatenate\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from tensorflow.keras.callbacks import Callback, ModelCheckpoint, EarlyStopping\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from tqdm import tqdm\n",
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:45.806415Z",
     "iopub.status.busy": "2021-03-21T04:01:45.805791Z",
     "iopub.status.idle": "2021-03-21T04:01:45.843270Z",
     "shell.execute_reply": "2021-03-21T04:01:45.843640Z"
    },
    "papermill": {
     "duration": 0.057056,
     "end_time": "2021-03-21T04:01:45.843788",
     "exception": false,
     "start_time": "2021-03-21T04:01:45.786732",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "img_path = '../input/severstal-steel-defect-detection/test_images/004f40c73.jpg'\n",
    "img = cv2.imread(img_path, cv2.IMREAD_GRAYSCALE)\n",
    "img = img.astype(np.float32) / 255.\n",
    "img = np.expand_dims(img, axis=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:45.880563Z",
     "iopub.status.busy": "2021-03-21T04:01:45.880049Z",
     "iopub.status.idle": "2021-03-21T04:01:45.883017Z",
     "shell.execute_reply": "2021-03-21T04:01:45.883408Z"
    },
    "papermill": {
     "duration": 0.024584,
     "end_time": "2021-03-21T04:01:45.883524",
     "exception": false,
     "start_time": "2021-03-21T04:01:45.858940",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(256, 1600, 1)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.shape(img)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "papermill": {
     "duration": 0.015294,
     "end_time": "2021-03-21T04:01:45.914337",
     "exception": false,
     "start_time": "2021-03-21T04:01:45.899043",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# load dataframe with train labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:45.949734Z",
     "iopub.status.busy": "2021-03-21T04:01:45.949237Z",
     "iopub.status.idle": "2021-03-21T04:01:46.359394Z",
     "shell.execute_reply": "2021-03-21T04:01:46.358493Z"
    },
    "papermill": {
     "duration": 0.429665,
     "end_time": "2021-03-21T04:01:46.359525",
     "exception": false,
     "start_time": "2021-03-21T04:01:45.929860",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(7095, 3)\n"
     ]
    }
   ],
   "source": [
    "train_df = pd.read_csv('../input/severstal-steel-defect-detection/train.csv')\n",
    "print(train_df.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:46.406858Z",
     "iopub.status.busy": "2021-03-21T04:01:46.406058Z",
     "iopub.status.idle": "2021-03-21T04:01:46.413045Z",
     "shell.execute_reply": "2021-03-21T04:01:46.412625Z"
    },
    "papermill": {
     "duration": 0.037534,
     "end_time": "2021-03-21T04:01:46.413157",
     "exception": false,
     "start_time": "2021-03-21T04:01:46.375623",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "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>ImageId</th>\n",
       "      <th>ClassId</th>\n",
       "      <th>EncodedPixels</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0002cc93b.jpg</td>\n",
       "      <td>1</td>\n",
       "      <td>29102 12 29346 24 29602 24 29858 24 30114 24 3...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0007a71bf.jpg</td>\n",
       "      <td>3</td>\n",
       "      <td>18661 28 18863 82 19091 110 19347 110 19603 11...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>000a4bcdd.jpg</td>\n",
       "      <td>1</td>\n",
       "      <td>37607 3 37858 8 38108 14 38359 20 38610 25 388...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>000f6bf48.jpg</td>\n",
       "      <td>4</td>\n",
       "      <td>131973 1 132228 4 132483 6 132738 8 132993 11 ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0014fce06.jpg</td>\n",
       "      <td>3</td>\n",
       "      <td>229501 11 229741 33 229981 55 230221 77 230468...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         ImageId  ClassId                                      EncodedPixels\n",
       "0  0002cc93b.jpg        1  29102 12 29346 24 29602 24 29858 24 30114 24 3...\n",
       "1  0007a71bf.jpg        3  18661 28 18863 82 19091 110 19347 110 19603 11...\n",
       "2  000a4bcdd.jpg        1  37607 3 37858 8 38108 14 38359 20 38610 25 388...\n",
       "3  000f6bf48.jpg        4  131973 1 132228 4 132483 6 132738 8 132993 11 ...\n",
       "4  0014fce06.jpg        3  229501 11 229741 33 229981 55 230221 77 230468..."
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.head(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:46.458257Z",
     "iopub.status.busy": "2021-03-21T04:01:46.457552Z",
     "iopub.status.idle": "2021-03-21T04:01:46.461192Z",
     "shell.execute_reply": "2021-03-21T04:01:46.461599Z"
    },
    "papermill": {
     "duration": 0.03222,
     "end_time": "2021-03-21T04:01:46.461719",
     "exception": false,
     "start_time": "2021-03-21T04:01:46.429499",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 7095 entries, 0 to 7094\n",
      "Data columns (total 3 columns):\n",
      " #   Column         Non-Null Count  Dtype \n",
      "---  ------         --------------  ----- \n",
      " 0   ImageId        7095 non-null   object\n",
      " 1   ClassId        7095 non-null   int64 \n",
      " 2   EncodedPixels  7095 non-null   object\n",
      "dtypes: int64(1), object(2)\n",
      "memory usage: 166.4+ KB\n"
     ]
    }
   ],
   "source": [
    "train_df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:46.500474Z",
     "iopub.status.busy": "2021-03-21T04:01:46.499882Z",
     "iopub.status.idle": "2021-03-21T04:01:46.503003Z",
     "shell.execute_reply": "2021-03-21T04:01:46.503387Z"
    },
    "papermill": {
     "duration": 0.025157,
     "end_time": "2021-03-21T04:01:46.503511",
     "exception": false,
     "start_time": "2021-03-21T04:01:46.478354",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[1, 3, 4, 2]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(train_df['ClassId'].unique())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:46.542109Z",
     "iopub.status.busy": "2021-03-21T04:01:46.541316Z",
     "iopub.status.idle": "2021-03-21T04:01:46.559244Z",
     "shell.execute_reply": "2021-03-21T04:01:46.558662Z"
    },
    "papermill": {
     "duration": 0.039333,
     "end_time": "2021-03-21T04:01:46.559383",
     "exception": false,
     "start_time": "2021-03-21T04:01:46.520050",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(6666, 2)\n"
     ]
    },
    {
     "data": {
      "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>ImageId</th>\n",
       "      <th>Num_ClassId</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>6253</th>\n",
       "      <td>ef24da2ba.jpg</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5740</th>\n",
       "      <td>db4867ee8.jpg</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3435</th>\n",
       "      <td>84865746c.jpg</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2203</th>\n",
       "      <td>5665ada1b.jpg</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6149</th>\n",
       "      <td>eb44ad79d.jpg</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            ImageId  Num_ClassId\n",
       "6253  ef24da2ba.jpg            3\n",
       "5740  db4867ee8.jpg            3\n",
       "3435  84865746c.jpg            2\n",
       "2203  5665ada1b.jpg            2\n",
       "6149  eb44ad79d.jpg            2"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mask_count_df = train_df.groupby('ImageId')[\"ClassId\"].count().reset_index().rename(columns = {\"ClassId\":\"Num_ClassId\"})\n",
    "mask_count_df.sort_values('Num_ClassId', ascending=False, inplace=True)\n",
    "print(mask_count_df.shape)\n",
    "mask_count_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "papermill": {
     "duration": 0.017188,
     "end_time": "2021-03-21T04:01:46.594177",
     "exception": false,
     "start_time": "2021-03-21T04:01:46.576989",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Keras U-Net"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "papermill": {
     "duration": 0.017118,
     "end_time": "2021-03-21T04:01:46.628593",
     "exception": false,
     "start_time": "2021-03-21T04:01:46.611475",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Utility Functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:46.672371Z",
     "iopub.status.busy": "2021-03-21T04:01:46.670453Z",
     "iopub.status.idle": "2021-03-21T04:01:46.674552Z",
     "shell.execute_reply": "2021-03-21T04:01:46.674941Z"
    },
    "papermill": {
     "duration": 0.029261,
     "end_time": "2021-03-21T04:01:46.675061",
     "exception": false,
     "start_time": "2021-03-21T04:01:46.645800",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def mask2rle(img):\n",
    "    '''\n",
    "    img: numpy array, 1 - mask, 0 - background\n",
    "    Returns run length as string formated\n",
    "    '''\n",
    "    pixels= img.T.flatten()\n",
    "    pixels = np.concatenate([[0], pixels, [0]])\n",
    "    runs = np.where(pixels[1:] != pixels[:-1])[0] + 1\n",
    "    runs[1::2] -= runs[::2]\n",
    "    return ' '.join(str(x) for x in runs)\n",
    "\n",
    "def rle2mask(rle, input_shape):\n",
    "    width, height = input_shape[:2]\n",
    "    \n",
    "    mask= np.zeros( width*height ).astype(np.uint8)\n",
    "    \n",
    "    array = np.asarray([int(x) for x in rle.split()])\n",
    "    starts = array[0::2]\n",
    "    starts-=1\n",
    "    lengths = array[1::2]\n",
    "\n",
    "    current_position = 0\n",
    "    for index, start in enumerate(starts):\n",
    "        mask[int(start):int(start+lengths[index])] = 1\n",
    "        current_position += lengths[index]\n",
    "        \n",
    "    return mask.reshape(height, width).T\n",
    "\n",
    "\n",
    "def build_rles(masks):\n",
    "    width, height, depth = masks.shape\n",
    "    \n",
    "    rles = [mask2rle(masks[:, :, i])\n",
    "            for i in range(depth)]\n",
    "    \n",
    "    return rles"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "papermill": {
     "duration": 0.017032,
     "end_time": "2021-03-21T04:01:46.709210",
     "exception": false,
     "start_time": "2021-03-21T04:01:46.692178",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Data Generator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:46.749317Z",
     "iopub.status.busy": "2021-03-21T04:01:46.748673Z",
     "iopub.status.idle": "2021-03-21T04:01:47.812694Z",
     "shell.execute_reply": "2021-03-21T04:01:47.814114Z"
    },
    "papermill": {
     "duration": 1.087786,
     "end_time": "2021-03-21T04:01:47.814373",
     "exception": false,
     "start_time": "2021-03-21T04:01:46.726587",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from albumentations import (\n",
    "    HorizontalFlip, VerticalFlip, IAAPerspective, ShiftScaleRotate, CLAHE, RandomRotate90,\n",
    "    Transpose, ShiftScaleRotate, Blur, OpticalDistortion, GridDistortion, HueSaturationValue,\n",
    "    IAAAdditiveGaussianNoise, GaussNoise, MotionBlur, MedianBlur, IAAPiecewiseAffine,\n",
    "    IAASharpen, IAAEmboss, RandomContrast, RandomBrightness, Flip, OneOf, Compose, RandomGamma, Rotate,IAAAffine\n",
    ")\n",
    "aug_null = <YOUR_CODE_HERE>([])\n",
    "aug = Compose([ \n",
    "    Blur(p=0.5, blur_limit=2),\n",
    "    #IAAAffine(p=0.5, shear=5),\n",
    "    HorizontalFlip(p=0.5),              \n",
    "    Rotate(limit=5, p=0.3),\n",
    "    VerticalFlip(p=0.5), \n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:47.887272Z",
     "iopub.status.busy": "2021-03-21T04:01:47.886409Z",
     "iopub.status.idle": "2021-03-21T04:01:47.907275Z",
     "shell.execute_reply": "2021-03-21T04:01:47.908291Z"
    },
    "papermill": {
     "duration": 0.063222,
     "end_time": "2021-03-21T04:01:47.908482",
     "exception": false,
     "start_time": "2021-03-21T04:01:47.845260",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "class DataGenerator(keras.utils.Sequence):\n",
    "    'Generates data for Keras'\n",
    "    def __init__(self, list_IDs, df, target_df=None, mode='fit',\n",
    "                 base_path='../input/severstal-steel-defect-detection/train_images',\n",
    "                 batch_size=32, dim=(256, 1600), n_channels=1,\n",
    "                 n_classes=4, random_state=2021, shuffle=True,\n",
    "                aug=aug_null):\n",
    "        self.dim = dim\n",
    "        self.batch_size = batch_size\n",
    "        self.df = df\n",
    "        self.mode = mode\n",
    "        self.base_path = base_path\n",
    "        self.target_df = target_df\n",
    "        self.list_IDs = list_IDs\n",
    "        self.n_channels = n_channels\n",
    "        self.n_classes = n_classes\n",
    "        self.shuffle = shuffle\n",
    "        self.random_state = random_state\n",
    "        self.aug = aug\n",
    "        \n",
    "        self.on_epoch_end()\n",
    "\n",
    "    def __len__(self):\n",
    "        'Denotes the number of batches per epoch'\n",
    "        return int(np.floor(len(self.list_IDs) / self.batch_size))\n",
    "\n",
    "    def __getitem__(self, index):\n",
    "        'Generate one batch of data'\n",
    "        # Generate indexes of the batch\n",
    "        indexes = self.indexes[index*self.batch_size:(index+1)*self.batch_size]\n",
    "\n",
    "        # Find list of IDs\n",
    "        list_IDs_batch = [self.list_IDs[k] for k in indexes]\n",
    "        \n",
    "        X = self.__generate_X(list_IDs_batch)\n",
    "        if self.mode == 'fit':\n",
    "            y = self.__generate_y(list_IDs_batch)\n",
    "            #print(np.shape(y))\n",
    "            #print(len(X))\n",
    "            \n",
    "            for img in range(len(X)):\n",
    "                augmented = self.aug(<YOUR_CODE_HERE>=X[img], <YOUR_CODE_HERE>=y[img] )\n",
    "                \n",
    "                X[img] = augmented['image']\n",
    "                y[img] = augmented['mask'] \n",
    "            return X, y\n",
    "        \n",
    "        elif self.mode == 'predict':\n",
    "            return X\n",
    "\n",
    "        else:\n",
    "            raise AttributeError('The mode parameter should be set to \"fit\" or \"predict\".')\n",
    "        \n",
    "    def on_epoch_end(self):\n",
    "        'Updates indexes after each epoch'\n",
    "        self.indexes = np.arange(len(self.list_IDs))\n",
    "        if self.shuffle == True:\n",
    "            np.random.seed(self.random_state)\n",
    "            np.random.shuffle(self.indexes)\n",
    "    \n",
    "    def __generate_X(self, list_IDs_batch):\n",
    "        'Generates data containing batch_size samples'\n",
    "        # Initialization\n",
    "        X = np.empty((self.batch_size, *self.dim, self.n_channels))\n",
    "        \n",
    "        # Generate data\n",
    "        for i, ID in enumerate(list_IDs_batch):\n",
    "            im_name = self.df['ImageId'].iloc[ID]\n",
    "            img_path = f\"{self.base_path}/{im_name}\"\n",
    "            img = self.__load_grayscale(img_path)\n",
    "            \n",
    "            # Store samples\n",
    "            X[i,] = img\n",
    "\n",
    "        return X\n",
    "    \n",
    "    def __generate_y(self, list_IDs_batch):\n",
    "        y = np.empty((self.batch_size, *self.dim, self.n_classes), dtype=int)\n",
    "        \n",
    "        for i, ID in enumerate(list_IDs_batch):\n",
    "            #print()\n",
    "            #print(\"i\",i)\n",
    "            #print(\"ID\",ID)\n",
    "            im_name = self.df['ImageId'].iloc[ID]\n",
    "            image_df = self.target_df[self.target_df['ImageId'] == im_name].copy().reset_index()\n",
    "            #print(\"image_df:\",image_df)\n",
    "\n",
    "            masks = np.zeros((*self.dim, self.n_classes))\n",
    "            \n",
    "            for j in range(len(image_df)):\n",
    "                #print(j)\n",
    "                rle = image_df.loc[j,'EncodedPixels']\n",
    "                cls =  image_df.loc[j,'ClassId']\n",
    "                masks[:, :, cls-1] = rle2mask(rle, self.dim)\n",
    "    \n",
    "            y[i, ] = masks\n",
    "\n",
    "        return y\n",
    "    \n",
    "    def __load_grayscale(self, img_path):\n",
    "        img = cv2.imread(img_path, cv2.IMREAD_GRAYSCALE)\n",
    "        img = img.astype(np.float32) / 255.\n",
    "        img = np.expand_dims(img, axis=-1)\n",
    "\n",
    "        return img\n",
    "    \n",
    "    def __load_rgb(self, img_path):\n",
    "        img = cv2.imread(img_path)\n",
    "        img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n",
    "        img = img.astype(np.float32) / 255.\n",
    "\n",
    "        return img"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:47.981927Z",
     "iopub.status.busy": "2021-03-21T04:01:47.979962Z",
     "iopub.status.idle": "2021-03-21T04:01:47.982569Z",
     "shell.execute_reply": "2021-03-21T04:01:47.983006Z"
    },
    "papermill": {
     "duration": 0.044014,
     "end_time": "2021-03-21T04:01:47.983142",
     "exception": false,
     "start_time": "2021-03-21T04:01:47.939128",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "BATCH_SIZE = 16\n",
    "\n",
    "train_idx, val_idx = train_test_split(\n",
    "    mask_count_df.index, \n",
    "    random_state=2021, \n",
    "    test_size=0.2\n",
    ")\n",
    "\n",
    "train_generator = DataGenerator(\n",
    "    train_idx, \n",
    "    df=mask_count_df,\n",
    "    target_df=train_df,\n",
    "    batch_size=BATCH_SIZE, \n",
    "    n_classes=4,\n",
    "    aug=aug \n",
    ")\n",
    "\n",
    "val_generator = DataGenerator(\n",
    "    val_idx, \n",
    "    df=mask_count_df,\n",
    "    target_df=train_df,\n",
    "    batch_size=BATCH_SIZE, \n",
    "    n_classes=4\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "papermill": {
     "duration": 0.018027,
     "end_time": "2021-03-21T04:01:48.018891",
     "exception": false,
     "start_time": "2021-03-21T04:01:48.000864",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:48.058609Z",
     "iopub.status.busy": "2021-03-21T04:01:48.058020Z",
     "iopub.status.idle": "2021-03-21T04:01:48.061949Z",
     "shell.execute_reply": "2021-03-21T04:01:48.061383Z"
    },
    "papermill": {
     "duration": 0.025661,
     "end_time": "2021-03-21T04:01:48.062109",
     "exception": false,
     "start_time": "2021-03-21T04:01:48.036448",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def dice_coef(y_true, y_pred, smooth=1):\n",
    "    y_true_f = K.flatten(y_true)\n",
    "    y_pred_f = K.flatten(y_pred)\n",
    "    intersection = K.sum(y_true_f * y_pred_f)\n",
    "    return (2. * intersection + smooth) / (K.sum(y_true_f) + K.sum(y_pred_f) + smooth)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:48.116114Z",
     "iopub.status.busy": "2021-03-21T04:01:48.104769Z",
     "iopub.status.idle": "2021-03-21T04:01:48.118258Z",
     "shell.execute_reply": "2021-03-21T04:01:48.118624Z"
    },
    "papermill": {
     "duration": 0.038856,
     "end_time": "2021-03-21T04:01:48.118742",
     "exception": false,
     "start_time": "2021-03-21T04:01:48.079886",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def build_model(input_shape):\n",
    "    inputs = Input(input_shape)\n",
    "\n",
    "    c1 = Conv2D(8, (3, 3), activation='relu', padding='same') (inputs)\n",
    "    c1 = Conv2D(8, (3, 3), activation='relu', padding='same') (c1)\n",
    "    p1 = MaxPooling2D((2, 2)) (c1)\n",
    "\n",
    "    c2 = Conv2D(16, (3, 3), activation='relu', padding='same') (p1)\n",
    "    c2 = Conv2D(16, (3, 3), activation='relu', padding='same') (c2)\n",
    "    p2 = MaxPooling2D((2, 2)) (c2)\n",
    "\n",
    "    c3 = Conv2D(32, (3, 3), activation='relu', padding='same') (p2)\n",
    "    c3 = Conv2D(32, (3, 3), activation='relu', padding='same') (c3)\n",
    "    p3 = MaxPooling2D((2, 2)) (c3)\n",
    "\n",
    "    c4 = Conv2D(64, (3, 3), activation='relu', padding='same') (p3)\n",
    "    c4 = Conv2D(64, (3, 3), activation='relu', padding='same') (c4)\n",
    "    p4 = MaxPooling2D(pool_size=(2, 2)) (c4)\n",
    "\n",
    "    c5 = Conv2D(64, (3, 3), activation='relu', padding='same') (p4)\n",
    "    c5 = Conv2D(64, (3, 3), activation='relu', padding='same') (c5)\n",
    "    p5 = MaxPooling2D(pool_size=(2, 2)) (c5)\n",
    "\n",
    "    c55 = Conv2D(128, (3, 3), activation='relu', padding='same') (p5)\n",
    "    c55 = Conv2D(128, (3, 3), activation='relu', padding='same') (c55)\n",
    "\n",
    "    u6 = Conv2DTranspose(64, (2, 2), strides=(2, 2), padding='same') (c55)\n",
    "    u6 = concatenate([u6, c5])\n",
    "    c6 = Conv2D(64, (3, 3), activation='relu', padding='same') (u6)\n",
    "    c6 = Conv2D(64, (3, 3), activation='relu', padding='same') (c6)\n",
    "\n",
    "    u71 = Conv2DTranspose(32, (2, 2), strides=(2, 2), padding='same') (c6)\n",
    "    u71 = concatenate([u71, c4])\n",
    "    c71 = Conv2D(32, (3, 3), activation='relu', padding='same') (u71)\n",
    "    c61 = Conv2D(32, (3, 3), activation='relu', padding='same') (c71)\n",
    "\n",
    "    u7 = Conv2DTranspose(32, (2, 2), strides=(2, 2), padding='same') (c61)\n",
    "    u7 = concatenate([u7, c3])\n",
    "    c7 = Conv2D(32, (3, 3), activation='relu', padding='same') (u7)\n",
    "    c7 = Conv2D(32, (3, 3), activation='relu', padding='same') (c7)\n",
    "\n",
    "    u8 = Conv2DTranspose(16, (2, 2), strides=(2, 2), padding='same') (c7)\n",
    "    u8 = concatenate([u8, c2])\n",
    "    c8 = Conv2D(16, (3, 3), activation='relu', padding='same') (u8)\n",
    "    c8 = Conv2D(16, (3, 3), activation='relu', padding='same') (c8)\n",
    "\n",
    "    u9 = Conv2DTranspose(8, (2, 2), strides=(2, 2), padding='same') (c8)\n",
    "    u9 = concatenate([u9, c1], axis=3)\n",
    "    c9 = Conv2D(8, (3, 3), activation='relu', padding='same') (u9)\n",
    "    c9 = Conv2D(8, (3, 3), activation='relu', padding='same') (c9)\n",
    "\n",
    "    outputs = Conv2D(4, (1, 1), activation='sigmoid') (c9)\n",
    "\n",
    "    model = Model(inputs=[inputs], outputs=[outputs])\n",
    "    model.compile(optimizer='adam', loss='binary_crossentropy', metrics=[dice_coef])\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:48.157106Z",
     "iopub.status.busy": "2021-03-21T04:01:48.156583Z",
     "iopub.status.idle": "2021-03-21T04:01:51.381769Z",
     "shell.execute_reply": "2021-03-21T04:01:51.382957Z"
    },
    "papermill": {
     "duration": 3.247118,
     "end_time": "2021-03-21T04:01:51.383168",
     "exception": false,
     "start_time": "2021-03-21T04:01:48.136050",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_1 (InputLayer)            [(None, 256, 1600, 1 0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d (Conv2D)                 (None, 256, 1600, 8) 80          input_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_1 (Conv2D)               (None, 256, 1600, 8) 584         conv2d[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D)    (None, 128, 800, 8)  0           conv2d_1[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_2 (Conv2D)               (None, 128, 800, 16) 1168        max_pooling2d[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_3 (Conv2D)               (None, 128, 800, 16) 2320        conv2d_2[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2D)  (None, 64, 400, 16)  0           conv2d_3[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_4 (Conv2D)               (None, 64, 400, 32)  4640        max_pooling2d_1[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_5 (Conv2D)               (None, 64, 400, 32)  9248        conv2d_4[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2D)  (None, 32, 200, 32)  0           conv2d_5[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_6 (Conv2D)               (None, 32, 200, 64)  18496       max_pooling2d_2[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_7 (Conv2D)               (None, 32, 200, 64)  36928       conv2d_6[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2D)  (None, 16, 100, 64)  0           conv2d_7[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_8 (Conv2D)               (None, 16, 100, 64)  36928       max_pooling2d_3[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_9 (Conv2D)               (None, 16, 100, 64)  36928       conv2d_8[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2D)  (None, 8, 50, 64)    0           conv2d_9[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_10 (Conv2D)              (None, 8, 50, 128)   73856       max_pooling2d_4[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_11 (Conv2D)              (None, 8, 50, 128)   147584      conv2d_10[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_transpose (Conv2DTranspo (None, 16, 100, 64)  32832       conv2d_11[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate (Concatenate)       (None, 16, 100, 128) 0           conv2d_transpose[0][0]           \n",
      "                                                                 conv2d_9[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_12 (Conv2D)              (None, 16, 100, 64)  73792       concatenate[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_13 (Conv2D)              (None, 16, 100, 64)  36928       conv2d_12[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_transpose_1 (Conv2DTrans (None, 32, 200, 32)  8224        conv2d_13[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_1 (Concatenate)     (None, 32, 200, 96)  0           conv2d_transpose_1[0][0]         \n",
      "                                                                 conv2d_7[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_14 (Conv2D)              (None, 32, 200, 32)  27680       concatenate_1[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_15 (Conv2D)              (None, 32, 200, 32)  9248        conv2d_14[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_transpose_2 (Conv2DTrans (None, 64, 400, 32)  4128        conv2d_15[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_2 (Concatenate)     (None, 64, 400, 64)  0           conv2d_transpose_2[0][0]         \n",
      "                                                                 conv2d_5[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_16 (Conv2D)              (None, 64, 400, 32)  18464       concatenate_2[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_17 (Conv2D)              (None, 64, 400, 32)  9248        conv2d_16[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_transpose_3 (Conv2DTrans (None, 128, 800, 16) 2064        conv2d_17[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_3 (Concatenate)     (None, 128, 800, 32) 0           conv2d_transpose_3[0][0]         \n",
      "                                                                 conv2d_3[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_18 (Conv2D)              (None, 128, 800, 16) 4624        concatenate_3[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_19 (Conv2D)              (None, 128, 800, 16) 2320        conv2d_18[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_transpose_4 (Conv2DTrans (None, 256, 1600, 8) 520         conv2d_19[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_4 (Concatenate)     (None, 256, 1600, 16 0           conv2d_transpose_4[0][0]         \n",
      "                                                                 conv2d_1[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_20 (Conv2D)              (None, 256, 1600, 8) 1160        concatenate_4[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_21 (Conv2D)              (None, 256, 1600, 8) 584         conv2d_20[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_22 (Conv2D)              (None, 256, 1600, 4) 36          conv2d_21[0][0]                  \n",
      "==================================================================================================\n",
      "Total params: 600,612\n",
      "Trainable params: 600,612\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = build_model((256, 1600, 1))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-21T04:01:51.436388Z",
     "iopub.status.busy": "2021-03-21T04:01:51.435797Z",
     "iopub.status.idle": "2021-03-21T10:21:30.771058Z",
     "shell.execute_reply": "2021-03-21T10:21:30.770374Z"
    },
    "papermill": {
     "duration": 22779.359296,
     "end_time": "2021-03-21T10:21:30.771228",
     "exception": false,
     "start_time": "2021-03-21T04:01:51.411932",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "333/333 - 295s - loss: 0.0972 - dice_coef: 0.0509 - val_loss: 0.0568 - val_dice_coef: 0.0956\n",
      "\n",
      "Epoch 00001: val_loss improved from inf to 0.05677, saving model to model.h5\n",
      "Epoch 2/100\n",
      "333/333 - 256s - loss: 0.0480 - dice_coef: 0.2048 - val_loss: 0.0411 - val_dice_coef: 0.2804\n",
      "\n",
      "Epoch 00002: val_loss improved from 0.05677 to 0.04114, saving model to model.h5\n",
      "Epoch 3/100\n",
      "333/333 - 260s - loss: 0.0404 - dice_coef: 0.3016 - val_loss: 0.0386 - val_dice_coef: 0.3187\n",
      "\n",
      "Epoch 00003: val_loss improved from 0.04114 to 0.03864, saving model to model.h5\n",
      "Epoch 4/100\n",
      "333/333 - 252s - loss: 0.0369 - dice_coef: 0.3557 - val_loss: 0.0373 - val_dice_coef: 0.4065\n",
      "\n",
      "Epoch 00004: val_loss improved from 0.03864 to 0.03731, saving model to model.h5\n",
      "Epoch 5/100\n",
      "333/333 - 260s - loss: 0.0352 - dice_coef: 0.3801 - val_loss: 0.0355 - val_dice_coef: 0.3305\n",
      "\n",
      "Epoch 00005: val_loss improved from 0.03731 to 0.03549, saving model to model.h5\n",
      "Epoch 6/100\n",
      "333/333 - 260s - loss: 0.0346 - dice_coef: 0.3890 - val_loss: 0.0342 - val_dice_coef: 0.4036\n",
      "\n",
      "Epoch 00006: val_loss improved from 0.03549 to 0.03419, saving model to model.h5\n",
      "Epoch 7/100\n",
      "333/333 - 263s - loss: 0.0340 - dice_coef: 0.3959 - val_loss: 0.0340 - val_dice_coef: 0.4235\n",
      "\n",
      "Epoch 00007: val_loss improved from 0.03419 to 0.03396, saving model to model.h5\n",
      "Epoch 8/100\n",
      "333/333 - 261s - loss: 0.0329 - dice_coef: 0.4128 - val_loss: 0.0320 - val_dice_coef: 0.4135\n",
      "\n",
      "Epoch 00008: val_loss improved from 0.03396 to 0.03196, saving model to model.h5\n",
      "Epoch 9/100\n",
      "333/333 - 261s - loss: 0.0325 - dice_coef: 0.4176 - val_loss: 0.0318 - val_dice_coef: 0.4095\n",
      "\n",
      "Epoch 00009: val_loss improved from 0.03196 to 0.03175, saving model to model.h5\n",
      "Epoch 10/100\n",
      "333/333 - 259s - loss: 0.0312 - dice_coef: 0.4318 - val_loss: 0.0332 - val_dice_coef: 0.3916\n",
      "\n",
      "Epoch 00010: val_loss did not improve from 0.03175\n",
      "Epoch 11/100\n",
      "333/333 - 259s - loss: 0.0310 - dice_coef: 0.4357 - val_loss: 0.0319 - val_dice_coef: 0.3790\n",
      "\n",
      "Epoch 00011: val_loss did not improve from 0.03175\n",
      "Epoch 12/100\n",
      "333/333 - 258s - loss: 0.0302 - dice_coef: 0.4473 - val_loss: 0.0318 - val_dice_coef: 0.4190\n",
      "\n",
      "Epoch 00012: val_loss did not improve from 0.03175\n",
      "Epoch 13/100\n",
      "333/333 - 257s - loss: 0.0298 - dice_coef: 0.4548 - val_loss: 0.0302 - val_dice_coef: 0.4237\n",
      "\n",
      "Epoch 00013: val_loss improved from 0.03175 to 0.03016, saving model to model.h5\n",
      "Epoch 14/100\n",
      "333/333 - 252s - loss: 0.0293 - dice_coef: 0.4655 - val_loss: 0.0361 - val_dice_coef: 0.4312\n",
      "\n",
      "Epoch 00014: val_loss did not improve from 0.03016\n",
      "Epoch 15/100\n",
      "333/333 - 251s - loss: 0.0286 - dice_coef: 0.4728 - val_loss: 0.0279 - val_dice_coef: 0.4541\n",
      "\n",
      "Epoch 00015: val_loss improved from 0.03016 to 0.02794, saving model to model.h5\n",
      "Epoch 16/100\n",
      "333/333 - 261s - loss: 0.0281 - dice_coef: 0.4821 - val_loss: 0.0281 - val_dice_coef: 0.5166\n",
      "\n",
      "Epoch 00016: val_loss did not improve from 0.02794\n",
      "Epoch 17/100\n",
      "333/333 - 260s - loss: 0.0275 - dice_coef: 0.4910 - val_loss: 0.0301 - val_dice_coef: 0.4854\n",
      "\n",
      "Epoch 00017: val_loss did not improve from 0.02794\n",
      "Epoch 18/100\n",
      "333/333 - 259s - loss: 0.0273 - dice_coef: 0.4964 - val_loss: 0.0282 - val_dice_coef: 0.5189\n",
      "\n",
      "Epoch 00018: val_loss did not improve from 0.02794\n",
      "Epoch 19/100\n",
      "333/333 - 260s - loss: 0.0266 - dice_coef: 0.5048 - val_loss: 0.0276 - val_dice_coef: 0.4944\n",
      "\n",
      "Epoch 00019: val_loss improved from 0.02794 to 0.02760, saving model to model.h5\n",
      "Epoch 20/100\n",
      "333/333 - 259s - loss: 0.0272 - dice_coef: 0.4962 - val_loss: 0.0276 - val_dice_coef: 0.5432\n",
      "\n",
      "Epoch 00020: val_loss improved from 0.02760 to 0.02756, saving model to model.h5\n",
      "Epoch 21/100\n",
      "333/333 - 259s - loss: 0.0258 - dice_coef: 0.5166 - val_loss: 0.0262 - val_dice_coef: 0.5171\n",
      "\n",
      "Epoch 00021: val_loss improved from 0.02756 to 0.02616, saving model to model.h5\n",
      "Epoch 22/100\n",
      "333/333 - 259s - loss: 0.0260 - dice_coef: 0.5151 - val_loss: 0.0299 - val_dice_coef: 0.4938\n",
      "\n",
      "Epoch 00022: val_loss did not improve from 0.02616\n",
      "Epoch 23/100\n",
      "333/333 - 260s - loss: 0.0257 - dice_coef: 0.5229 - val_loss: 0.0312 - val_dice_coef: 0.4903\n",
      "\n",
      "Epoch 00023: val_loss did not improve from 0.02616\n",
      "Epoch 24/100\n",
      "333/333 - 261s - loss: 0.0253 - dice_coef: 0.5219 - val_loss: 0.0252 - val_dice_coef: 0.5618\n",
      "\n",
      "Epoch 00024: val_loss improved from 0.02616 to 0.02523, saving model to model.h5\n",
      "Epoch 25/100\n",
      "333/333 - 262s - loss: 0.0257 - dice_coef: 0.5186 - val_loss: 0.0297 - val_dice_coef: 0.4947\n",
      "\n",
      "Epoch 00025: val_loss did not improve from 0.02523\n",
      "Epoch 26/100\n",
      "333/333 - 261s - loss: 0.0256 - dice_coef: 0.5221 - val_loss: 0.0270 - val_dice_coef: 0.5169\n",
      "\n",
      "Epoch 00026: val_loss did not improve from 0.02523\n",
      "Epoch 27/100\n",
      "333/333 - 260s - loss: 0.0247 - dice_coef: 0.5336 - val_loss: 0.0253 - val_dice_coef: 0.5073\n",
      "\n",
      "Epoch 00027: val_loss did not improve from 0.02523\n",
      "Epoch 28/100\n",
      "333/333 - 262s - loss: 0.0244 - dice_coef: 0.5371 - val_loss: 0.0243 - val_dice_coef: 0.5654\n",
      "\n",
      "Epoch 00028: val_loss improved from 0.02523 to 0.02435, saving model to model.h5\n",
      "Epoch 29/100\n",
      "333/333 - 263s - loss: 0.0240 - dice_coef: 0.5434 - val_loss: 0.0244 - val_dice_coef: 0.5669\n",
      "\n",
      "Epoch 00029: val_loss did not improve from 0.02435\n",
      "Epoch 30/100\n",
      "333/333 - 262s - loss: 0.0240 - dice_coef: 0.5449 - val_loss: 0.0250 - val_dice_coef: 0.5501\n",
      "\n",
      "Epoch 00030: val_loss did not improve from 0.02435\n",
      "Epoch 31/100\n",
      "333/333 - 261s - loss: 0.0241 - dice_coef: 0.5433 - val_loss: 0.0247 - val_dice_coef: 0.5522\n",
      "\n",
      "Epoch 00031: val_loss did not improve from 0.02435\n",
      "Epoch 32/100\n",
      "333/333 - 261s - loss: 0.0237 - dice_coef: 0.5510 - val_loss: 0.0256 - val_dice_coef: 0.5324\n",
      "\n",
      "Epoch 00032: val_loss did not improve from 0.02435\n",
      "Epoch 33/100\n",
      "333/333 - 261s - loss: 0.0234 - dice_coef: 0.5552 - val_loss: 0.0238 - val_dice_coef: 0.5381\n",
      "\n",
      "Epoch 00033: val_loss improved from 0.02435 to 0.02381, saving model to model.h5\n",
      "Epoch 34/100\n",
      "333/333 - 261s - loss: 0.0231 - dice_coef: 0.5586 - val_loss: 0.0239 - val_dice_coef: 0.5762\n",
      "\n",
      "Epoch 00034: val_loss did not improve from 0.02381\n",
      "Epoch 35/100\n",
      "333/333 - 260s - loss: 0.0235 - dice_coef: 0.5553 - val_loss: 0.0237 - val_dice_coef: 0.5685\n",
      "\n",
      "Epoch 00035: val_loss improved from 0.02381 to 0.02367, saving model to model.h5\n",
      "Epoch 36/100\n",
      "333/333 - 253s - loss: 0.0230 - dice_coef: 0.5610 - val_loss: 0.0234 - val_dice_coef: 0.5754\n",
      "\n",
      "Epoch 00036: val_loss improved from 0.02367 to 0.02344, saving model to model.h5\n",
      "Epoch 37/100\n",
      "333/333 - 260s - loss: 0.0231 - dice_coef: 0.5600 - val_loss: 0.0238 - val_dice_coef: 0.5595\n",
      "\n",
      "Epoch 00037: val_loss did not improve from 0.02344\n",
      "Epoch 38/100\n",
      "333/333 - 260s - loss: 0.0225 - dice_coef: 0.5685 - val_loss: 0.0243 - val_dice_coef: 0.5751\n",
      "\n",
      "Epoch 00038: val_loss did not improve from 0.02344\n",
      "Epoch 39/100\n",
      "333/333 - 259s - loss: 0.0226 - dice_coef: 0.5661 - val_loss: 0.0241 - val_dice_coef: 0.5551\n",
      "\n",
      "Epoch 00039: val_loss did not improve from 0.02344\n",
      "Epoch 40/100\n",
      "333/333 - 261s - loss: 0.0221 - dice_coef: 0.5752 - val_loss: 0.0238 - val_dice_coef: 0.5643\n",
      "\n",
      "Epoch 00040: val_loss did not improve from 0.02344\n",
      "Epoch 41/100\n",
      "333/333 - 262s - loss: 0.0227 - dice_coef: 0.5681 - val_loss: 0.0230 - val_dice_coef: 0.5655\n",
      "\n",
      "Epoch 00041: val_loss improved from 0.02344 to 0.02297, saving model to model.h5\n",
      "Epoch 42/100\n",
      "333/333 - 262s - loss: 0.0220 - dice_coef: 0.5771 - val_loss: 0.0238 - val_dice_coef: 0.5714\n",
      "\n",
      "Epoch 00042: val_loss did not improve from 0.02297\n",
      "Epoch 43/100\n",
      "333/333 - 261s - loss: 0.0227 - dice_coef: 0.5674 - val_loss: 0.0234 - val_dice_coef: 0.5798\n",
      "\n",
      "Epoch 00043: val_loss did not improve from 0.02297\n",
      "Epoch 44/100\n",
      "333/333 - 262s - loss: 0.0223 - dice_coef: 0.5724 - val_loss: 0.0229 - val_dice_coef: 0.5543\n",
      "\n",
      "Epoch 00044: val_loss improved from 0.02297 to 0.02286, saving model to model.h5\n",
      "Epoch 45/100\n",
      "333/333 - 261s - loss: 0.0223 - dice_coef: 0.5705 - val_loss: 0.0239 - val_dice_coef: 0.5724\n",
      "\n",
      "Epoch 00045: val_loss did not improve from 0.02286\n",
      "Epoch 46/100\n",
      "333/333 - 262s - loss: 0.0215 - dice_coef: 0.5852 - val_loss: 0.0234 - val_dice_coef: 0.5843\n",
      "\n",
      "Epoch 00046: val_loss did not improve from 0.02286\n",
      "Epoch 47/100\n",
      "333/333 - 263s - loss: 0.0217 - dice_coef: 0.5817 - val_loss: 0.0231 - val_dice_coef: 0.5812\n",
      "\n",
      "Epoch 00047: val_loss did not improve from 0.02286\n",
      "Epoch 48/100\n",
      "333/333 - 262s - loss: 0.0214 - dice_coef: 0.5879 - val_loss: 0.0237 - val_dice_coef: 0.5684\n",
      "\n",
      "Epoch 00048: val_loss did not improve from 0.02286\n",
      "Epoch 49/100\n",
      "333/333 - 263s - loss: 0.0215 - dice_coef: 0.5855 - val_loss: 0.0230 - val_dice_coef: 0.5762\n",
      "\n",
      "Epoch 00049: val_loss did not improve from 0.02286\n",
      "Epoch 50/100\n",
      "333/333 - 261s - loss: 0.0216 - dice_coef: 0.5846 - val_loss: 0.0232 - val_dice_coef: 0.5622\n",
      "\n",
      "Epoch 00050: val_loss did not improve from 0.02286\n",
      "Epoch 51/100\n",
      "333/333 - 262s - loss: 0.0214 - dice_coef: 0.5900 - val_loss: 0.0220 - val_dice_coef: 0.5836\n",
      "\n",
      "Epoch 00051: val_loss improved from 0.02286 to 0.02204, saving model to model.h5\n",
      "Epoch 52/100\n",
      "333/333 - 262s - loss: 0.0211 - dice_coef: 0.5927 - val_loss: 0.0235 - val_dice_coef: 0.5648\n",
      "\n",
      "Epoch 00052: val_loss did not improve from 0.02204\n",
      "Epoch 53/100\n",
      "333/333 - 261s - loss: 0.0211 - dice_coef: 0.5920 - val_loss: 0.0224 - val_dice_coef: 0.5769\n",
      "\n",
      "Epoch 00053: val_loss did not improve from 0.02204\n",
      "Epoch 54/100\n",
      "333/333 - 262s - loss: 0.0209 - dice_coef: 0.5950 - val_loss: 0.0240 - val_dice_coef: 0.5731\n",
      "\n",
      "Epoch 00054: val_loss did not improve from 0.02204\n",
      "Epoch 55/100\n",
      "333/333 - 261s - loss: 0.0211 - dice_coef: 0.5933 - val_loss: 0.0241 - val_dice_coef: 0.5447\n",
      "\n",
      "Epoch 00055: val_loss did not improve from 0.02204\n",
      "Epoch 56/100\n",
      "333/333 - 268s - loss: 0.0211 - dice_coef: 0.5899 - val_loss: 0.0229 - val_dice_coef: 0.5829\n",
      "\n",
      "Epoch 00056: val_loss did not improve from 0.02204\n",
      "Epoch 57/100\n",
      "333/333 - 265s - loss: 0.0205 - dice_coef: 0.6025 - val_loss: 0.0219 - val_dice_coef: 0.5915\n",
      "\n",
      "Epoch 00057: val_loss improved from 0.02204 to 0.02188, saving model to model.h5\n",
      "Epoch 58/100\n",
      "333/333 - 265s - loss: 0.0205 - dice_coef: 0.6002 - val_loss: 0.0252 - val_dice_coef: 0.5649\n",
      "\n",
      "Epoch 00058: val_loss did not improve from 0.02188\n",
      "Epoch 59/100\n",
      "333/333 - 262s - loss: 0.0206 - dice_coef: 0.5997 - val_loss: 0.0222 - val_dice_coef: 0.5784\n",
      "\n",
      "Epoch 00059: val_loss did not improve from 0.02188\n",
      "Epoch 60/100\n",
      "333/333 - 265s - loss: 0.0209 - dice_coef: 0.5964 - val_loss: 0.0237 - val_dice_coef: 0.5989\n",
      "\n",
      "Epoch 00060: val_loss did not improve from 0.02188\n",
      "Epoch 61/100\n",
      "333/333 - 265s - loss: 0.0203 - dice_coef: 0.6070 - val_loss: 0.0276 - val_dice_coef: 0.5523\n",
      "\n",
      "Epoch 00061: val_loss did not improve from 0.02188\n",
      "Epoch 62/100\n",
      "333/333 - 264s - loss: 0.0203 - dice_coef: 0.6030 - val_loss: 0.0223 - val_dice_coef: 0.6027\n",
      "\n",
      "Epoch 00062: val_loss did not improve from 0.02188\n",
      "Epoch 63/100\n",
      "333/333 - 262s - loss: 0.0204 - dice_coef: 0.6030 - val_loss: 0.0234 - val_dice_coef: 0.5876\n",
      "\n",
      "Epoch 00063: val_loss did not improve from 0.02188\n",
      "Epoch 64/100\n",
      "333/333 - 263s - loss: 0.0200 - dice_coef: 0.6128 - val_loss: 0.0231 - val_dice_coef: 0.5682\n",
      "\n",
      "Epoch 00064: val_loss did not improve from 0.02188\n",
      "Epoch 65/100\n",
      "333/333 - 262s - loss: 0.0197 - dice_coef: 0.6145 - val_loss: 0.0222 - val_dice_coef: 0.5963\n",
      "\n",
      "Epoch 00065: val_loss did not improve from 0.02188\n",
      "Epoch 66/100\n",
      "333/333 - 263s - loss: 0.0202 - dice_coef: 0.6065 - val_loss: 0.0222 - val_dice_coef: 0.6170\n",
      "\n",
      "Epoch 00066: val_loss did not improve from 0.02188\n",
      "Epoch 67/100\n",
      "333/333 - 263s - loss: 0.0197 - dice_coef: 0.6150 - val_loss: 0.0228 - val_dice_coef: 0.6037\n",
      "\n",
      "Epoch 00067: val_loss did not improve from 0.02188\n",
      "Epoch 68/100\n",
      "333/333 - 266s - loss: 0.0201 - dice_coef: 0.6088 - val_loss: 0.0224 - val_dice_coef: 0.6035\n",
      "\n",
      "Epoch 00068: val_loss did not improve from 0.02188\n",
      "Epoch 69/100\n",
      "333/333 - 264s - loss: 0.0193 - dice_coef: 0.6219 - val_loss: 0.0221 - val_dice_coef: 0.6086\n",
      "\n",
      "Epoch 00069: val_loss did not improve from 0.02188\n",
      "Epoch 70/100\n",
      "333/333 - 264s - loss: 0.0195 - dice_coef: 0.6197 - val_loss: 0.0226 - val_dice_coef: 0.6060\n",
      "\n",
      "Epoch 00070: val_loss did not improve from 0.02188\n",
      "Epoch 71/100\n",
      "333/333 - 264s - loss: 0.0193 - dice_coef: 0.6200 - val_loss: 0.0230 - val_dice_coef: 0.6240\n",
      "\n",
      "Epoch 00071: val_loss did not improve from 0.02188\n",
      "Epoch 72/100\n",
      "333/333 - 265s - loss: 0.0193 - dice_coef: 0.6207 - val_loss: 0.0222 - val_dice_coef: 0.5903\n",
      "\n",
      "Epoch 00072: val_loss did not improve from 0.02188\n",
      "Epoch 73/100\n",
      "333/333 - 264s - loss: 0.0191 - dice_coef: 0.6257 - val_loss: 0.0240 - val_dice_coef: 0.6059\n",
      "\n",
      "Epoch 00073: val_loss did not improve from 0.02188\n",
      "Epoch 74/100\n",
      "333/333 - 263s - loss: 0.0191 - dice_coef: 0.6273 - val_loss: 0.0316 - val_dice_coef: 0.5410\n",
      "\n",
      "Epoch 00074: val_loss did not improve from 0.02188\n",
      "Epoch 75/100\n",
      "333/333 - 263s - loss: 0.0193 - dice_coef: 0.6211 - val_loss: 0.0223 - val_dice_coef: 0.6059\n",
      "\n",
      "Epoch 00075: val_loss did not improve from 0.02188\n",
      "Epoch 76/100\n",
      "333/333 - 261s - loss: 0.0195 - dice_coef: 0.6188 - val_loss: 0.0222 - val_dice_coef: 0.6069\n",
      "\n",
      "Epoch 00076: val_loss did not improve from 0.02188\n",
      "Epoch 77/100\n",
      "333/333 - 262s - loss: 0.0188 - dice_coef: 0.6306 - val_loss: 0.0223 - val_dice_coef: 0.6033\n",
      "\n",
      "Epoch 00077: val_loss did not improve from 0.02188\n",
      "Epoch 78/100\n",
      "333/333 - 261s - loss: 0.0186 - dice_coef: 0.6315 - val_loss: 0.0225 - val_dice_coef: 0.6170\n",
      "\n",
      "Epoch 00078: val_loss did not improve from 0.02188\n",
      "Epoch 79/100\n",
      "333/333 - 262s - loss: 0.0189 - dice_coef: 0.6287 - val_loss: 0.0231 - val_dice_coef: 0.5987\n",
      "\n",
      "Epoch 00079: val_loss did not improve from 0.02188\n",
      "Epoch 80/100\n",
      "333/333 - 263s - loss: 0.0185 - dice_coef: 0.6336 - val_loss: 0.0226 - val_dice_coef: 0.5961\n",
      "\n",
      "Epoch 00080: val_loss did not improve from 0.02188\n",
      "Epoch 81/100\n",
      "333/333 - 263s - loss: 0.0185 - dice_coef: 0.6355 - val_loss: 0.0235 - val_dice_coef: 0.5946\n",
      "\n",
      "Epoch 00081: val_loss did not improve from 0.02188\n",
      "Epoch 82/100\n",
      "333/333 - 264s - loss: 0.0184 - dice_coef: 0.6374 - val_loss: 0.0219 - val_dice_coef: 0.6248\n",
      "\n",
      "Epoch 00082: val_loss did not improve from 0.02188\n",
      "Epoch 83/100\n",
      "333/333 - 265s - loss: 0.0182 - dice_coef: 0.6416 - val_loss: 0.0231 - val_dice_coef: 0.6091\n",
      "\n",
      "Epoch 00083: val_loss did not improve from 0.02188\n",
      "Epoch 84/100\n",
      "333/333 - 266s - loss: 0.0192 - dice_coef: 0.6270 - val_loss: 0.0228 - val_dice_coef: 0.6090\n",
      "\n",
      "Epoch 00084: val_loss did not improve from 0.02188\n",
      "Epoch 85/100\n",
      "333/333 - 265s - loss: 0.0183 - dice_coef: 0.6395 - val_loss: 0.0228 - val_dice_coef: 0.6076\n",
      "\n",
      "Epoch 00085: val_loss did not improve from 0.02188\n",
      "Epoch 86/100\n",
      "333/333 - 261s - loss: 0.0180 - dice_coef: 0.6429 - val_loss: 0.0261 - val_dice_coef: 0.5717\n",
      "\n",
      "Epoch 00086: val_loss did not improve from 0.02188\n",
      "Epoch 87/100\n",
      "333/333 - 263s - loss: 0.0182 - dice_coef: 0.6412 - val_loss: 0.0231 - val_dice_coef: 0.5983\n",
      "\n",
      "Epoch 00087: val_loss did not improve from 0.02188\n",
      "Epoch 00087: early stopping\n"
     ]
    }
   ],
   "source": [
    "mc = ModelCheckpoint(\n",
    "    'model.h5', \n",
    "    monitor='val_loss', \n",
    "    verbose=1, \n",
    "    save_best_only=True, \n",
    "    save_weights_only=False,\n",
    "    mode='auto'\n",
    ")\n",
    "\n",
    "es = EarlyStopping(monitor='val_loss', mode='min', verbose=2, patience=30)\n",
    "\n",
    "history = model.fit(\n",
    "    train_generator,\n",
    "    validation_data=val_generator,\n",
    "    callbacks=[mc,es],\n",
    "    verbose=2, \n",
    "    epochs=100\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "papermill": {
     "duration": 0.103021,
     "end_time": "2021-03-21T10:21:30.978810",
     "exception": false,
     "start_time": "2021-03-21T10:21:30.875789",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-21T10:21:31.178059Z",
     "iopub.status.busy": "2021-03-21T10:21:31.177142Z",
     "iopub.status.idle": "2021-03-21T10:21:31.179774Z",
     "shell.execute_reply": "2021-03-21T10:21:31.179347Z"
    },
    "papermill": {
     "duration": 0.092268,
     "end_time": "2021-03-21T10:21:31.179891",
     "exception": false,
     "start_time": "2021-03-21T10:21:31.087623",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def history_plot(model_id, history1):\n",
    "\tplt.figure(figsize=(14,16)) \n",
    "\t# plot loss\n",
    "\tplt.subplot(2, 1, 1)\n",
    "\tplt.title('Cross Entropy Loss - ' + model_id, fontsize = 12)\n",
    "\tplt.plot(history1.history['loss'], color='blue', label='train base')\n",
    "\tplt.plot(history1.history['val_loss'], color='cyan', label='val base')\n",
    "\tplt.xlabel(\"Epochs\",fontsize = 12)\n",
    "\tplt.ylabel(\"Loss\", fontsize = 12)\n",
    "\tplt.legend(loc='upper right')\n",
    "\n",
    "\t# plot accuracy\n",
    "\tplt.subplot(2, 1, 2)\n",
    "\tplt.title('Classification Accuracy ' + model_id, fontsize = 10)\n",
    "\tplt.plot(history1.history['dice_coef'], color='blue', label='train base')\n",
    "\tplt.plot(history1.history['val_dice_coef'], color='cyan', label='val base')\n",
    "\tplt.xlabel(\"Epochs\", fontsize = 12)\n",
    "\tplt.ylabel(\"dice_coef\", fontsize = 12)\n",
    "\tplt.legend(loc='lower right')\n",
    "\n",
    "\tplt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-21T10:21:31.310709Z",
     "iopub.status.busy": "2021-03-21T10:21:31.309937Z",
     "iopub.status.idle": "2021-03-21T10:21:31.313723Z",
     "shell.execute_reply": "2021-03-21T10:21:31.313319Z"
    },
    "papermill": {
     "duration": 0.070173,
     "end_time": "2021-03-21T10:21:31.313850",
     "exception": false,
     "start_time": "2021-03-21T10:21:31.243677",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['loss', 'dice_coef', 'val_loss', 'val_dice_coef'])"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "history.history.keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-21T10:21:31.451558Z",
     "iopub.status.busy": "2021-03-21T10:21:31.450717Z",
     "iopub.status.idle": "2021-03-21T10:21:31.808381Z",
     "shell.execute_reply": "2021-03-21T10:21:31.808820Z"
    },
    "papermill": {
     "duration": 0.431375,
     "end_time": "2021-03-21T10:21:31.808973",
     "exception": false,
     "start_time": "2021-03-21T10:21:31.377598",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1008x1152 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "history_plot('UNET Model', history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-21T10:21:31.969501Z",
     "iopub.status.busy": "2021-03-21T10:21:31.962642Z",
     "iopub.status.idle": "2021-03-21T10:21:31.989091Z",
     "shell.execute_reply": "2021-03-21T10:21:31.988265Z"
    },
    "papermill": {
     "duration": 0.115094,
     "end_time": "2021-03-21T10:21:31.989212",
     "exception": false,
     "start_time": "2021-03-21T10:21:31.874118",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "model.load_weights('model.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-21T10:21:32.134229Z",
     "iopub.status.busy": "2021-03-21T10:21:32.133239Z",
     "iopub.status.idle": "2021-03-21T10:21:32.255226Z",
     "shell.execute_reply": "2021-03-21T10:21:32.254682Z"
    },
    "papermill": {
     "duration": 0.201413,
     "end_time": "2021-03-21T10:21:32.255356",
     "exception": false,
     "start_time": "2021-03-21T10:21:32.053943",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "model.save('severstal-segmentation_unetDataAugLoss.h5', save_format=\"h5\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.3"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 22800.618797,
   "end_time": "2021-03-21T10:21:35.457974",
   "environment_variables": {},
   "exception": null,
   "input_path": "__notebook__.ipynb",
   "output_path": "__notebook__.ipynb",
   "parameters": {},
   "start_time": "2021-03-21T04:01:34.839177",
   "version": "2.2.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}