encoder_decoder.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from tqdm import tqdm\n",
    "import tensorflow as tf\n",
    "import cv2\n",
    "from PIL import Image,ImageOps\n",
    "from PIL import ImageDraw \n",
    "import shutil\n",
    "import os\n",
    "from PIL import Image\n",
    "import pickle\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator,array_to_img,img_to_array\n",
    "from keras.applications.vgg19 import preprocess_input\n",
    "from tensorflow.keras.preprocessing.text import Tokenizer\n",
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "from sklearn.model_selection import train_test_split\n",
    "import time\n",
    "from tensorflow.keras.models import Model\n",
    "from tensorflow.keras.layers import Dense, LSTM, Input, Embedding, Conv2D, Concatenate, Flatten, Add, Dropout, GRU\n",
    "import random\n",
    "import datetime\n",
    "from nltk.translate.bleu_score import sentence_bleu"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'/Users/sashrikasurya/Documents/Medical-Report-Generator'"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dirname = os.path.dirname(os.path.realpath('__file__'))\n",
    "dirname"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = pd.read_csv(dirname+\"/data/traindata.csv\",nrows=1000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "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>Unnamed: 0</th>\n",
       "      <th>patient_id</th>\n",
       "      <th>image1_id</th>\n",
       "      <th>image2_id</th>\n",
       "      <th>report</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>CXR3080_IM-1440</td>\n",
       "      <td>CXR3080_IM-1440-1001</td>\n",
       "      <td>CXR3080_IM-1440-1002</td>\n",
       "      <td>startseq the cavity and the left upper lobe ha...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>CXR2721_IM-1183</td>\n",
       "      <td>CXR2721_IM-1183-1001</td>\n",
       "      <td>CXR2721_IM-1183-2001</td>\n",
       "      <td>startseq heart size at the upper limits of nor...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>CXR2160_IM-0778</td>\n",
       "      <td>CXR2160_IM-0778-1001</td>\n",
       "      <td>CXR2160_IM-0778-2001</td>\n",
       "      <td>startseq the cardiomediastinal silhouette is w...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>CXR3586_IM-1764</td>\n",
       "      <td>CXR3586_IM-1764-1001</td>\n",
       "      <td>CXR3586_IM-1764-2001</td>\n",
       "      <td>startseq the heart is mildly enlarged the medi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>CXR3139_IM-1476</td>\n",
       "      <td>CXR3139_IM-1476-1001</td>\n",
       "      <td>CXR3139_IM-1476-2001</td>\n",
       "      <td>startseq no focal areas of consolidation no pl...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0       patient_id             image1_id             image2_id  \\\n",
       "0           0  CXR3080_IM-1440  CXR3080_IM-1440-1001  CXR3080_IM-1440-1002   \n",
       "1           1  CXR2721_IM-1183  CXR2721_IM-1183-1001  CXR2721_IM-1183-2001   \n",
       "2           2  CXR2160_IM-0778  CXR2160_IM-0778-1001  CXR2160_IM-0778-2001   \n",
       "3           3  CXR3586_IM-1764  CXR3586_IM-1764-1001  CXR3586_IM-1764-2001   \n",
       "4           4  CXR3139_IM-1476  CXR3139_IM-1476-1001  CXR3139_IM-1476-2001   \n",
       "\n",
       "                                              report  \n",
       "0  startseq the cavity and the left upper lobe ha...  \n",
       "1  startseq heart size at the upper limits of nor...  \n",
       "2  startseq the cardiomediastinal silhouette is w...  \n",
       "3  startseq the heart is mildly enlarged the medi...  \n",
       "4  startseq no focal areas of consolidation no pl...  "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "test = pd.read_csv(dirname+\"/data/testdata.csv\",nrows=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "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>Unnamed: 0</th>\n",
       "      <th>patient_id</th>\n",
       "      <th>image1_id</th>\n",
       "      <th>image2_id</th>\n",
       "      <th>report</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>CXR1306_IM-0200</td>\n",
       "      <td>CXR1306_IM-0200-2001</td>\n",
       "      <td>CXR1306_IM-0200-3001</td>\n",
       "      <td>startseq the  examination consists of frontal ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>CXR1786_IM-0512</td>\n",
       "      <td>CXR1786_IM-0512-1001</td>\n",
       "      <td>CXR1786_IM-0512-2001</td>\n",
       "      <td>startseq the lungs are clear bilaterally speci...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>CXR2072_IM-0706</td>\n",
       "      <td>CXR2072_IM-0706-1001</td>\n",
       "      <td>CXR2072_IM-0706-2001</td>\n",
       "      <td>startseq stable nonenlarged cardiomediastinal ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>CXR467_IM-2095</td>\n",
       "      <td>CXR467_IM-2095-2001</td>\n",
       "      <td>CXR467_IM-2095-2001</td>\n",
       "      <td>startseq lungs are clear bilaterally cardiac a...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>CXR3087_IM-1444</td>\n",
       "      <td>CXR3087_IM-1444-1001</td>\n",
       "      <td>CXR3087_IM-1444-2001</td>\n",
       "      <td>startseq the heart is enlarged the mediastinal...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0       patient_id             image1_id             image2_id  \\\n",
       "0           0  CXR1306_IM-0200  CXR1306_IM-0200-2001  CXR1306_IM-0200-3001   \n",
       "1           1  CXR1786_IM-0512  CXR1786_IM-0512-1001  CXR1786_IM-0512-2001   \n",
       "2           2  CXR2072_IM-0706  CXR2072_IM-0706-1001  CXR2072_IM-0706-2001   \n",
       "3           3   CXR467_IM-2095   CXR467_IM-2095-2001   CXR467_IM-2095-2001   \n",
       "4           4  CXR3087_IM-1444  CXR3087_IM-1444-1001  CXR3087_IM-1444-2001   \n",
       "\n",
       "                                              report  \n",
       "0  startseq the  examination consists of frontal ...  \n",
       "1  startseq the lungs are clear bilaterally speci...  \n",
       "2  startseq stable nonenlarged cardiomediastinal ...  \n",
       "3  startseq lungs are clear bilaterally cardiac a...  \n",
       "4  startseq the heart is enlarged the mediastinal...  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "train.drop('Unnamed: 0',axis='columns',inplace=True)\n",
    "test.drop('Unnamed: 0',axis='columns',inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "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>patient_id</th>\n",
       "      <th>image1_id</th>\n",
       "      <th>image2_id</th>\n",
       "      <th>report</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>CXR3080_IM-1440</td>\n",
       "      <td>CXR3080_IM-1440-1001</td>\n",
       "      <td>CXR3080_IM-1440-1002</td>\n",
       "      <td>startseq the cavity and the left upper lobe ha...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>CXR2721_IM-1183</td>\n",
       "      <td>CXR2721_IM-1183-1001</td>\n",
       "      <td>CXR2721_IM-1183-2001</td>\n",
       "      <td>startseq heart size at the upper limits of nor...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>CXR2160_IM-0778</td>\n",
       "      <td>CXR2160_IM-0778-1001</td>\n",
       "      <td>CXR2160_IM-0778-2001</td>\n",
       "      <td>startseq the cardiomediastinal silhouette is w...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>CXR3586_IM-1764</td>\n",
       "      <td>CXR3586_IM-1764-1001</td>\n",
       "      <td>CXR3586_IM-1764-2001</td>\n",
       "      <td>startseq the heart is mildly enlarged the medi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>CXR3139_IM-1476</td>\n",
       "      <td>CXR3139_IM-1476-1001</td>\n",
       "      <td>CXR3139_IM-1476-2001</td>\n",
       "      <td>startseq no focal areas of consolidation no pl...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        patient_id             image1_id             image2_id  \\\n",
       "0  CXR3080_IM-1440  CXR3080_IM-1440-1001  CXR3080_IM-1440-1002   \n",
       "1  CXR2721_IM-1183  CXR2721_IM-1183-1001  CXR2721_IM-1183-2001   \n",
       "2  CXR2160_IM-0778  CXR2160_IM-0778-1001  CXR2160_IM-0778-2001   \n",
       "3  CXR3586_IM-1764  CXR3586_IM-1764-1001  CXR3586_IM-1764-2001   \n",
       "4  CXR3139_IM-1476  CXR3139_IM-1476-1001  CXR3139_IM-1476-2001   \n",
       "\n",
       "                                              report  \n",
       "0  startseq the cavity and the left upper lobe ha...  \n",
       "1  startseq heart size at the upper limits of nor...  \n",
       "2  startseq the cardiomediastinal silhouette is w...  \n",
       "3  startseq the heart is mildly enlarged the medi...  \n",
       "4  startseq no focal areas of consolidation no pl...  "
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "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>patient_id</th>\n",
       "      <th>image1_id</th>\n",
       "      <th>image2_id</th>\n",
       "      <th>report</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>CXR1306_IM-0200</td>\n",
       "      <td>CXR1306_IM-0200-2001</td>\n",
       "      <td>CXR1306_IM-0200-3001</td>\n",
       "      <td>startseq the  examination consists of frontal ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>CXR1786_IM-0512</td>\n",
       "      <td>CXR1786_IM-0512-1001</td>\n",
       "      <td>CXR1786_IM-0512-2001</td>\n",
       "      <td>startseq the lungs are clear bilaterally speci...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>CXR2072_IM-0706</td>\n",
       "      <td>CXR2072_IM-0706-1001</td>\n",
       "      <td>CXR2072_IM-0706-2001</td>\n",
       "      <td>startseq stable nonenlarged cardiomediastinal ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>CXR467_IM-2095</td>\n",
       "      <td>CXR467_IM-2095-2001</td>\n",
       "      <td>CXR467_IM-2095-2001</td>\n",
       "      <td>startseq lungs are clear bilaterally cardiac a...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>CXR3087_IM-1444</td>\n",
       "      <td>CXR3087_IM-1444-1001</td>\n",
       "      <td>CXR3087_IM-1444-2001</td>\n",
       "      <td>startseq the heart is enlarged the mediastinal...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        patient_id             image1_id             image2_id  \\\n",
       "0  CXR1306_IM-0200  CXR1306_IM-0200-2001  CXR1306_IM-0200-3001   \n",
       "1  CXR1786_IM-0512  CXR1786_IM-0512-1001  CXR1786_IM-0512-2001   \n",
       "2  CXR2072_IM-0706  CXR2072_IM-0706-1001  CXR2072_IM-0706-2001   \n",
       "3   CXR467_IM-2095   CXR467_IM-2095-2001   CXR467_IM-2095-2001   \n",
       "4  CXR3087_IM-1444  CXR3087_IM-1444-1001  CXR3087_IM-1444-2001   \n",
       "\n",
       "                                              report  \n",
       "0  startseq the  examination consists of frontal ...  \n",
       "1  startseq the lungs are clear bilaterally speci...  \n",
       "2  startseq stable nonenlarged cardiomediastinal ...  \n",
       "3  startseq lungs are clear bilaterally cardiac a...  \n",
       "4  startseq the heart is enlarged the mediastinal...  "
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((1000, 4), (200, 4))"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.shape, test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test = train['patient_id'], test['patient_id']\n",
    "y_train, y_test = train['report'], test['report']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((1000,), (1000,), (200,), (200,))"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape, y_train.shape, X_test.shape, y_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer = Tokenizer(filters='!\"#$%&()*+,-/:;<=>?@[\\\\]^_`{|}~\\t\\n')\n",
    "tokenizer.fit_on_texts(y_train.values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(dirname+'/features/list_train_features.pkl', 'rb') as f:\n",
    "    train_feat_list = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(dirname+'/features/list_test_features.pkl', 'rb') as f:\n",
    "    test_feat_list = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1000"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(train_feat_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "200"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(test_feat_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "f = open(dirname+'/glove.840B.300d.pkl','rb') # 300d glove vectors  \n",
    "glove_vectors = pickle.load(f)\n",
    "f.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "vocab_size = len(tokenizer.word_index.keys()) + 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "embedding_matrix = np.zeros((vocab_size,300))\n",
    "for word, i in tokenizer.word_index.items():\n",
    "    if word in glove_vectors.keys():\n",
    "        vec = glove_vectors[word]\n",
    "        embedding_matrix[i] = vec\n",
    "    else:\n",
    "        continue"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_size = 15"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_dataset = tf.data.Dataset.from_tensor_slices((train_feat_list,y_train))\n",
    "train_dataset = train_dataset.shuffle(300).batch(batch_size).prefetch(buffer_size=tf.data.experimental.AUTOTUNE)\n",
    "\n",
    "test_dataset = tf.data.Dataset.from_tensor_slices((test_feat_list,y_test))\n",
    "test_dataset = test_dataset.shuffle(300).batch(batch_size).prefetch(buffer_size=tf.data.experimental.AUTOTUNE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_data(images,reports_unpadded):\n",
    "    imgs = []\n",
    "    input_reports = []\n",
    "    output_reports = []\n",
    "    for i in range(len(images)):\n",
    "        sequence=[] \n",
    "        for rep in reports_unpadded[i].split(\" \"):\n",
    "            if rep in tokenizer.word_index.keys():\n",
    "                sequence.append(tokenizer.word_index[rep])\n",
    "        for j in range(1,len(sequence)):\n",
    "            in_seq = sequence[:j]          \n",
    "            out_seq = sequence[j]            \n",
    "            out_seq = tf.keras.utils.to_categorical(out_seq, num_classes=vocab_size)\n",
    "            imgs.append(images[i])            \n",
    "            input_reports.append(in_seq)            \n",
    "            output_reports.append(out_seq)\n",
    "        \n",
    "    return np.array(imgs), np.array(input_reports), np.array(output_reports)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_function = tf.keras.losses.CategoricalCrossentropy(from_logits=False, reduction='auto')\n",
    "def maskedLoss(y_true, y_pred):\n",
    "    mask = tf.math.logical_not(tf.math.equal(y_true, 0))\n",
    "    loss_ = loss_function(y_true, y_pred)\n",
    "    mask = tf.cast(mask, dtype=loss_.dtype)\n",
    "    loss_ = loss_*mask\n",
    "    loss_ = tf.reduce_mean(loss_)\n",
    "    return loss_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "max_len=80"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_7\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "text (InputLayer)               [(None, 80)]         0                                            \n",
      "__________________________________________________________________________________________________\n",
      "embedding_2 (Embedding)         (None, 80, 300)      321600      text[0][0]                       \n",
      "__________________________________________________________________________________________________\n",
      "LSTM1 (LSTM)                    (None, 80, 256)      570368      embedding_2[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "Images (InputLayer)             [(None, 2048)]       0                                            \n",
      "__________________________________________________________________________________________________\n",
      "LSTM2 (LSTM)                    (None, 512)          1574912     LSTM1[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "enc_dense (Dense)               (None, 512)          1049088     Images[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "dropout_4 (Dropout)             (None, 512)          0           LSTM2[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "add_2 (Add)                     (None, 512)          0           enc_dense[0][0]                  \n",
      "                                                                 dropout_4[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_7 (Dense)                 (None, 512)          262656      add_2[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "dropout_5 (Dropout)             (None, 512)          0           dense_7[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_8 (Dense)                 (None, 1072)         549936      dropout_5[0][0]                  \n",
      "==================================================================================================\n",
      "Total params: 4,328,560\n",
      "Trainable params: 4,006,960\n",
      "Non-trainable params: 321,600\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "input_1=Input(shape=(2048),name=\"Images\")\n",
    "encoder_out=Dense(512,activation=\"relu\",name=\"enc_dense\")(input_1)\n",
    "\n",
    "\n",
    "#decoder model\n",
    "input_text=Input(shape=(max_len),name=\"text\")\n",
    "\n",
    "embedding_out=tf.keras.layers.Embedding(input_dim=vocab_size,output_dim=300,input_length=max_len,mask_zero=True,trainable=False,weights=[embedding_matrix])(input_text)\n",
    "\n",
    "lstm_out1= LSTM(units=256, activation='tanh', recurrent_activation='sigmoid', use_bias=True, \n",
    "            kernel_initializer=tf.keras.initializers.glorot_uniform(seed=23),\n",
    "            recurrent_initializer=tf.keras.initializers.orthogonal(seed=7),\n",
    "            bias_initializer=tf.keras.initializers.zeros(), return_sequences=True, name=\"LSTM1\")(embedding_out)\n",
    "\n",
    "lstm_out2= LSTM(units=512, activation='tanh', recurrent_activation='sigmoid', use_bias=True, \n",
    "            kernel_initializer=tf.keras.initializers.glorot_uniform(seed=23),\n",
    "            recurrent_initializer=tf.keras.initializers.orthogonal(seed=7),\n",
    "            bias_initializer=tf.keras.initializers.zeros(), name=\"LSTM2\")(lstm_out1) \n",
    "\n",
    "x=Dropout(0.35)(lstm_out2)\n",
    "add=tf.keras.layers.Add()([encoder_out, x])\n",
    "  \n",
    "x=Dense(512,kernel_initializer=tf.keras.initializers.he_normal(seed=1),activation=\"relu\")(add)\n",
    "\n",
    "x1=Dropout(0.25)(x)\n",
    "\n",
    "x1=Dense(vocab_size,activation=\"softmax\")(x1)\n",
    "#encoder_decoder_model\n",
    "encoder_decoder=Model(inputs=[input_1,input_text],outputs=x1)\n",
    "encoder_decoder.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [],
   "source": [
    "encoder_decoder.compile(optimizer=\"Adam\", loss = maskedLoss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH :  1\n",
      "Training Loss: 0.004879826228275444\n"
     ]
    }
   ],
   "source": [
    "train_loss_list = []\n",
    "\n",
    "for epoch in range(1):\n",
    "    print('EPOCH : ',epoch+1)\n",
    "    batch_loss_train = 0\n",
    "    \n",
    "    for img, report in train_dataset:        \n",
    "        r1 = [word.decode('utf-8') for word in np.array(report)]  \n",
    "        img_input, rep_input, output_word = load_data(img.numpy(), r1)\n",
    "        rep_input = tf.keras.preprocessing.sequence.pad_sequences(rep_input, maxlen=80, padding='post')\n",
    "        img_input=tf.reshape(img_input,shape=(img_input.shape[0],img_input.shape[-1]))        \n",
    "        loss = encoder_decoder.train_on_batch([img_input, rep_input], output_word)        \n",
    "        batch_loss_train += loss\n",
    "\n",
    "    train_loss = batch_loss_train/(len(y_train)//15)\n",
    "    print('Training Loss: {}'.format(train_loss))\n",
    "    \n",
    "encoder_decoder.save_weights('encoder_decoder_epoch_'+ str(epoch+1) + '.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [],
   "source": [
    "def img_preprocess(img):\n",
    "    img=img_to_array(img)\n",
    "    img=preprocess_input(img)\n",
    "    img=cv2.resize(img,(224,224))\n",
    "    img=img/255.0\n",
    "    img=np.expand_dims(img, axis=0)\n",
    "    return img"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.applications import DenseNet121\n",
    "\n",
    "image_shape= (224,224,3)\n",
    "mod2=DenseNet121(include_top=False,input_shape=image_shape,pooling=\"avg\")\n",
    "las2=Dense(14,\"sigmoid\")(mod2.output)\n",
    "\n",
    "mod2=Model(inputs=mod2.input,outputs=las2)\n",
    "mod2.load_weights(dirname+\"/features/chexnet_weights.h5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [],
   "source": [
    "final_chexnet_model=Model(inputs=mod2.inputs,outputs=mod2.layers[-2].output,name=\"Chexnet_model\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [],
   "source": [
    "image_1= tf.keras.Input(shape=(224,224,3),name=\"image_1_features\")\n",
    "image_2= tf.keras.Input(shape=(224,224,3),name=\"image_2_features\")\n",
    "image_1_out=final_chexnet_model(image_1)\n",
    "image_2_out=final_chexnet_model(image_2)\n",
    "conc=tf.keras.layers.Concatenate(axis=-1)([image_1_out,image_2_out])\n",
    "feature_extraction_model=Model(inputs=[image_1,image_2],outputs=conc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [],
   "source": [
    "encoder_decoder.load_weights(dirname+'/encoder_decoder_epoch_1.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [],
   "source": [
    "def evaluation(image1,image2):\n",
    "    image_features=feature_extraction_model([image1,image2])\n",
    "    output_report=''\n",
    "    input_rep= 'startseq'\n",
    "    image_features=tf.reshape(image_features,shape=(-1,image_features.shape[-1]))\n",
    "    \n",
    "    for i in range(max_len):\n",
    "        input_tokens = [tokenizer.word_index[w] for w in input_rep.split()]\n",
    "        input_padded = tf.keras.preprocessing.sequence.pad_sequences([input_tokens],max_len, padding='post')\n",
    "        results = encoder_decoder.predict([image_features,input_padded])\n",
    "        arg = np.argmax(results[0]) \n",
    "        if tokenizer.index_word[arg]=='endseq':\n",
    "            output_report+=tokenizer.index_word[arg]+\" \"\n",
    "            break\n",
    "        else:\n",
    "            input_rep = input_rep + ' ' + tokenizer.index_word[arg]\n",
    "            output_report = output_report+tokenizer.index_word[arg]+\" \"\n",
    "    return output_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Actual Report:  startseq the  examination consists of frontal and lateral radiographs of the chest there are diminished lung volumes the cardiomediastinal contours are within normal limits pulmonary vascularity is within normal limits no focal consolidation pleural effusion or pneumothorax identified  scoliosis is unchanged visualized upper abdomen is grossly unremarkable endseq\n",
      "Generated Report:  the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the \n",
      "BLEU SCORE IS:  0.33437015248821106\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x158a5ce48>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "img1=Image.open(dirname+'/data/sample/CXR1009_IM-0010-1001'+'.png')\n",
    "img2=Image.open(dirname+'/data/sample/CXR1009_IM-0010-2001'+'.png')\n",
    "\n",
    "imshow(np.asarray(img1))\n",
    "img1=img_preprocess(img1)\n",
    "img2=img_preprocess(img2)\n",
    "result=evaluation(img1,img2) \n",
    "actual=y_test[0]\n",
    "\n",
    "print(\"Actual Report: \",actual)\n",
    "print(\"Generated Report: \",result) \n",
    "print(\"BLEU SCORE IS: \",sentence_bleu(actual,result))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "img1=Image.open(dirname+'/data/sample/CXR1026_IM-0021-2002'+'.png')\n",
    "img2=Image.open(dirname+'/data/sample/CXR1026_IM-0021-2002'+'.png')\n",
    "\n",
    "imshow(np.asarray(img1))\n",
    "img1=img_preprocess(img1)\n",
    "img2=img_preprocess(img2)\n",
    "result=evaluation(img1,img2) \n",
    "actual=y_test[4]\n",
    "print(\"Actual Report: \",actual)\n",
    "print(\"Generated Report: \",result) \n",
    "print(\"BLEU SCORE IS: \",sentence_bleu(actual,result))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "from matplotlib.pyplot import imshow"
   ]
  },
  {
   "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}