3_7_densenet169_with_pretrained_weights_3channel.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autotime"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time: 1.5 s\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "from matplotlib import pyplot as plt\n",
    "%matplotlib inline\n",
    "from keras.applications.densenet import DenseNet169\n",
    "from keras.utils import np_utils\n",
    "from keras.utils import to_categorical\n",
    "from keras.models import load_model\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras import layers\n",
    "from keras import optimizers\n",
    "from keras import metrics\n",
    "from keras import models\n",
    "\n",
    "from keras.utils import plot_model\n",
    "from IPython.display import SVG\n",
    "from keras.utils.vis_utils import model_to_dot\n",
    "from keras_tqdm import TQDMNotebookCallback\n",
    "from ipywidgets import IntProgress, HTML, HBox\n",
    "\n",
    "from pathlib import Path\n",
    "import pickle\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time: 740 µs\n"
     ]
    }
   ],
   "source": [
    "import logging\n",
    "\n",
    "log = logging.getLogger(__name__)\n",
    "logging.basicConfig(level=logging.INFO)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time: 1.43 s\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "from keras import backend as K\n",
    "config = tf.ConfigProto()\n",
    "config.gpu_options.allow_growth=True\n",
    "sess = tf.Session(config=config)\n",
    "K.set_session(sess)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time: 4.03 s\n"
     ]
    }
   ],
   "source": [
    "import utils # importing from our utils functions\n",
    "from PIL import Image, ImageOps"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:utils.utils:loading data/MURA-v1.1/x_train.pkl\n",
      "INFO:utils.utils:loading data/MURA-v1.1/y_train.pkl\n",
      "INFO:utils.utils:loading data/MURA-v1.1/x_valid.pkl\n",
      "INFO:utils.utils:loading data/MURA-v1.1/y_valid.pkl\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time: 7.98 s\n"
     ]
    }
   ],
   "source": [
    "x_train, y_train, x_test, y_test = utils.read_mura_pickle()\n",
    "x_train = utils.normalize_pixels(x_train)\n",
    "x_test = utils.normalize_pixels(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time: 16.2 s\n"
     ]
    }
   ],
   "source": [
    "x_train = np.stack([x_train, x_train, x_train], axis=3)\n",
    "x_test = np.stack([x_test, x_test, x_test], axis=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=RGB size=256x256 at 0x7F570014BBE0>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time: 30.1 ms\n"
     ]
    }
   ],
   "source": [
    "Image.fromarray((x_train[0] * 255).astype('uint8'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train Submodel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time: 1.77 ms\n"
     ]
    }
   ],
   "source": [
    "datagen = ImageDataGenerator(\n",
    "    rotation_range=45,\n",
    "    width_shift_range=0.2,\n",
    "    height_shift_range=0.2,\n",
    "    zoom_range=[1.0, 1.2],\n",
    "    fill_mode = 'constant',\n",
    "    cval=0,\n",
    "    horizontal_flip=True,\n",
    "    vertical_flip=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time: 21.8 s\n"
     ]
    }
   ],
   "source": [
    "datagen.fit(x_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time: 29.3 s\n"
     ]
    }
   ],
   "source": [
    "with tf.device('/GPU:0'):\n",
    "    densenet = DenseNet169(input_shape=(256, 256, 3), include_top=False)\n",
    "    for layer in densenet.layers:\n",
    "        layer.trainable = False\n",
    "    model = models.Model(input=densenet.input, output=layers.Dense(1, activation='sigmoid')(layers.Flatten()(densenet.output)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time: 1.44 ms\n"
     ]
    }
   ],
   "source": [
    "starting_epoch = 0\n",
    "train_round = 0\n",
    "batch_size = 16\n",
    "epochs_per_eval = 10\n",
    "epochs_per_save = 30\n",
    "total_epochs = 30\n",
    "# optimizer = 'adam'\n",
    "# lr = 0.001\n",
    "train_history = {'loss': [], 'binary_accuracy': [], 'learn_rate': []}\n",
    "val_per_image_history = {'accuracy': [], 'kappa': [], 'precision': [], 'recall': []}\n",
    "val_per_study_history = {'accuracy': [], 'kappa': [], 'precision': [], 'recall': []}\n",
    "model_name='3_7_shared_densenet169'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time: 15.6 ms\n"
     ]
    }
   ],
   "source": [
    "# clr = utils.clr.CyclicLR(mode='triangular2')\n",
    "# opt = optimizers.SGD(nesterov=True)\n",
    "opt = optimizers.Adam()"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "with tf.device('/GPU:0'):\n",
    "    model = models.load_model(f'./trained_models/{model_name}_{starting_epoch}.h5')"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "with open(f'./trained_models/{model_name}_train_history.pkl', 'rb') as pkl_file:\n",
    "    train_history = pickle.load(pkl_file)\n",
    "with open(f'./trained_models/{model_name}_val_per_image_history.pkl', 'rb') as pkl_file:\n",
    "    val_per_image_history = pickle.load(pkl_file)\n",
    "with open(f'./trained_models/{model_name}_val_per_study_history.pkl', 'rb') as pkl_file:\n",
    "    val_per_study_history = pickle.load(pkl_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:__main__:executing epochs 1-10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "dfc7248b366041b78c72847b324d9489",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Training', max=10), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 0', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 326s - loss: 4.1257 - binary_accuracy: 0.6513\n",
      "Epoch 2/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 1', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 330s - loss: 4.6200 - binary_accuracy: 0.6736\n",
      "Epoch 3/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 2', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 325s - loss: 4.6628 - binary_accuracy: 0.6833\n",
      "Epoch 4/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 3', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 319s - loss: 4.7079 - binary_accuracy: 0.6855\n",
      "Epoch 5/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 4', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 324s - loss: 4.7120 - binary_accuracy: 0.6889\n",
      "Epoch 6/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 5', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 330s - loss: 4.7296 - binary_accuracy: 0.6900\n",
      "Epoch 7/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 6', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 330s - loss: 4.6870 - binary_accuracy: 0.6945\n",
      "Epoch 8/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 7', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 325s - loss: 4.7701 - binary_accuracy: 0.6903\n",
      "Epoch 9/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 8', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 331s - loss: 4.7121 - binary_accuracy: 0.6957\n",
      "Epoch 10/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 9', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 329s - loss: 4.7137 - binary_accuracy: 0.6951\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:utils.mura_metrics:Accuracy: 0.6380982170785111\n",
      "INFO:utils.mura_metrics:Kappa: 0.25961315395838347\n",
      "INFO:utils.mura_metrics:Precision: 0.744429882044561\n",
      "INFO:utils.mura_metrics:Recall: 0.3712418300653595\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:utils.mura_metrics:Accuracy: 0.6513761467889908\n",
      "INFO:utils.mura_metrics:Kappa: 0.24938369971873345\n",
      "INFO:utils.mura_metrics:Precision: 0.8260869565217391\n",
      "INFO:utils.mura_metrics:Recall: 0.2825278810408922\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:__main__:Saving evaluation metrics history...\n",
      "INFO:__main__:executing epochs 11-20\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0cc0863d2bcc46eabe4c944793d097f1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Training', max=10), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 0', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 333s - loss: 4.9745 - binary_accuracy: 0.6822\n",
      "Epoch 2/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 1', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 328s - loss: 4.8310 - binary_accuracy: 0.6910\n",
      "Epoch 3/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 2', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 330s - loss: 4.7900 - binary_accuracy: 0.6946\n",
      "Epoch 4/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 3', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 323s - loss: 4.7498 - binary_accuracy: 0.6958\n",
      "Epoch 5/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 4', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 329s - loss: 4.7009 - binary_accuracy: 0.7000\n",
      "Epoch 6/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 5', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 315s - loss: 4.7410 - binary_accuracy: 0.6971\n",
      "Epoch 7/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 6', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 325s - loss: 4.7007 - binary_accuracy: 0.7002\n",
      "Epoch 8/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 7', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 311s - loss: 4.6946 - binary_accuracy: 0.7002\n",
      "Epoch 9/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 8', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 307s - loss: 4.7192 - binary_accuracy: 0.6996\n",
      "Epoch 10/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 9', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 322s - loss: 4.6968 - binary_accuracy: 0.7008\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:utils.mura_metrics:Accuracy: 0.5833593994369721\n",
      "INFO:utils.mura_metrics:Kappa: 0.13580180770350048\n",
      "INFO:utils.mura_metrics:Precision: 0.875\n",
      "INFO:utils.mura_metrics:Recall: 0.15098039215686274\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:utils.mura_metrics:Accuracy: 0.5904920767306089\n",
      "INFO:utils.mura_metrics:Kappa: 0.09649145693321327\n",
      "INFO:utils.mura_metrics:Precision: 0.9433962264150944\n",
      "INFO:utils.mura_metrics:Recall: 0.09293680297397769\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:__main__:Saving evaluation metrics history...\n",
      "INFO:__main__:executing epochs 21-30\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3c08f5190e6e42de89d18c92dc793a92",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Training', max=10), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 0', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 331s - loss: 4.8728 - binary_accuracy: 0.6908\n",
      "Epoch 2/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 1', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 315s - loss: 4.8103 - binary_accuracy: 0.6957\n",
      "Epoch 3/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 2', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 324s - loss: 4.7399 - binary_accuracy: 0.6989\n",
      "Epoch 4/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 3', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 317s - loss: 4.6763 - binary_accuracy: 0.7038\n",
      "Epoch 5/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 4', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 319s - loss: 4.7898 - binary_accuracy: 0.6960\n",
      "Epoch 6/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 5', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 324s - loss: 4.7423 - binary_accuracy: 0.6985\n",
      "Epoch 7/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 6', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 383s - loss: 4.6778 - binary_accuracy: 0.7040\n",
      "Epoch 8/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 7', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 330s - loss: 4.7509 - binary_accuracy: 0.6990\n",
      "Epoch 9/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 8', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 313s - loss: 4.6399 - binary_accuracy: 0.7054\n",
      "Epoch 10/10\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch 9', max=2300), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 323s - loss: 4.7171 - binary_accuracy: 0.7015\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:utils.mura_metrics:Accuracy: 0.6312167657178605\n",
      "INFO:utils.mura_metrics:Kappa: 0.2436682959404334\n",
      "INFO:utils.mura_metrics:Precision: 0.7607726597325408\n",
      "INFO:utils.mura_metrics:Recall: 0.33464052287581697\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:utils.mura_metrics:Accuracy: 0.6438698915763136\n",
      "INFO:utils.mura_metrics:Kappa: 0.23081326256041518\n",
      "INFO:utils.mura_metrics:Precision: 0.8323353293413174\n",
      "INFO:utils.mura_metrics:Recall: 0.258364312267658\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:__main__:Saving Model to trained_models/3_7_shared_densenet169_30.h5...\n",
      "INFO:__main__:Saving evaluation metrics history...\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time: 2h 45min 37s\n"
     ]
    }
   ],
   "source": [
    "while (train_round+1) * epochs_per_eval + starting_epoch <= total_epochs:\n",
    "    log.info(f'executing epochs {train_round * epochs_per_eval + 1 + starting_epoch}-{(train_round+1) * epochs_per_eval + starting_epoch}')\n",
    "    \n",
    "#     if train_round * epochs_per_eval + 1 + starting_epoch >= 10:\n",
    "#         if np.mean(train_history['loss'][-10:-5]) < np.mean(train_history['loss'][-5:]): # if no improvements in 5 epochs\n",
    "#             if optimizer == 'adam':\n",
    "#                 log.info(f'switching to SGD with lr={lr}')\n",
    "#                 optimizer = 'sgd'\n",
    "#             else:\n",
    "#                 log.info(f'reducing SGD learn rate from {lr} to {lr * 0.8}')\n",
    "#                 lr = lr * 0.8\n",
    "#             opt = optimizers.SGD(lr=lr, momentum=0.9, nesterov=True)\n",
    "#         else:\n",
    "#             if optimizer == 'adam':\n",
    "#                 opt = optimizers.Adam(lr=lr)\n",
    "#             else:\n",
    "#                 opt = optimizers.SGD(lr=lr, momentum=0.9, nesterov=True)\n",
    "#     else:\n",
    "#         opt = optimizers.Adam(lr=lr)\n",
    "\n",
    "    model.compile(\n",
    "        loss='binary_crossentropy',\n",
    "        optimizer=opt,\n",
    "        metrics=[metrics.binary_accuracy])\n",
    "    \n",
    "    model.fit_generator(datagen.flow(x_train, y_train, batch_size=batch_size),\n",
    "                           steps_per_epoch = int(np.round(len(x_train) / batch_size)), \n",
    "                           epochs=epochs_per_eval, verbose=2,\n",
    "                           callbacks=[TQDMNotebookCallback()])\n",
    "                           # callbacks=[clr, TQDMNotebookCallback()])\n",
    "    \n",
    "    # train_history['learn_rate'].extend(clr.history['lr'])\n",
    "    train_history['loss'].extend(model.history.history['loss'])\n",
    "    train_history['binary_accuracy'].extend(model.history.history['binary_accuracy'])\n",
    "    \n",
    "    y_valid_hat = model.predict(x_test)\n",
    "    true_label = np.round(y_test)\n",
    "    pred_label = np.round(y_valid_hat)\n",
    "    evaluate = utils.MURAMetrics(true_label, pred_label)\n",
    "    \n",
    "    per_image_metrics = evaluate.report_by_image()\n",
    "    for key in per_image_metrics:\n",
    "        val_per_image_history[key].append(per_image_metrics[key])\n",
    "    \n",
    "    per_study_metrics = evaluate.report_by_study()\n",
    "    for key in per_image_metrics:\n",
    "        val_per_study_history[key].append(per_study_metrics[key])\n",
    "    \n",
    "    cur_epoch = (train_round + 1) * epochs_per_eval\n",
    "    if cur_epoch % epochs_per_save == 0:\n",
    "        log.info(f'Saving Model to trained_models/{model_name}_{cur_epoch + starting_epoch}.h5...')\n",
    "        model.save(f'./trained_models/{model_name}_{cur_epoch + starting_epoch}.h5')\n",
    "    \n",
    "    log.info('Saving evaluation metrics history...')\n",
    "    with open(f'./trained_models/{model_name}_train_history.pkl', 'wb') as pkl_file:\n",
    "        pickle.dump(train_history, pkl_file)\n",
    "    with open(f'./trained_models/{model_name}_val_per_image_history.pkl', 'wb') as pkl_file:\n",
    "        pickle.dump(val_per_image_history, pkl_file)\n",
    "    with open(f'./trained_models/{model_name}_val_per_study_history.pkl', 'wb') as pkl_file:\n",
    "        pickle.dump(val_per_study_history, pkl_file)\n",
    "    \n",
    "    train_round += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}