Malarial_cell_classifier.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cv2\n",
    "import numpy as np \n",
    "import os\n",
    "import matplotlib.pyplot as plt \n",
    "import pandas as pd \n",
    "import tensorflow as tf\n",
    "import tensorflow.keras as keras \n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from sklearn.model_selection import train_test_split\n",
    "from tensorflow.keras.layers import Conv2D, Dropout, Dense, MaxPool2D, Flatten, BatchNormalization\n",
    "from tensorflow.keras import callbacks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "unhealthy = ['Data/Parasitized/' + i for i in os.listdir('Data/Parasitized/')]\n",
    "healthy = ['Data/Uninfected/' + i for i in os.listdir('Data/Uninfected/')]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_features = unhealthy[:12402]\n",
    "train_features.extend(healthy[:12402])\n",
    "train_labels = ['unhealthy']*12402\n",
    "train_labels.extend(['healthy']*12402)\n",
    "\n",
    "valid_features = unhealthy[12402:13091]\n",
    "valid_features.extend(healthy[12402:13091])\n",
    "valid_labels = ['unhealthy']*689\n",
    "valid_labels.extend(['healthy']*689)\n",
    "\n",
    "test_features = unhealthy[13091:]\n",
    "test_features.extend(healthy[13091:])\n",
    "test_labels = ['unhealthy']*689\n",
    "test_labels.extend(['healthy']*689)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df = pd.DataFrame({'path':train_features, 'label':train_labels})\n",
    "valid_df = pd.DataFrame({'path':valid_features, 'label':test_labels})\n",
    "test_df = pd.DataFrame({'path':test_features, 'label':valid_labels})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 24804 validated image filenames belonging to 2 classes.\n",
      "Found 1378 validated image filenames belonging to 2 classes.\n",
      "Found 1376 validated image filenames belonging to 2 classes.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "D:\\Anaconda\\envs\\tf2\\lib\\site-packages\\keras_preprocessing\\image\\dataframe_iterator.py:273: UserWarning: Found 2 invalid image filename(s) in x_col=\"path\". These filename(s) will be ignored.\n",
      "  .format(n_invalid, x_col)\n"
     ]
    }
   ],
   "source": [
    "datagen = ImageDataGenerator(rescale=1./255)\n",
    "\n",
    "train_gen = datagen.flow_from_dataframe(train_df,\n",
    "                                       x_col='path',\n",
    "                                       y_col='label',\n",
    "                                       batch_size=32,\n",
    "                                       shuffle=True,\n",
    "                                       seed=1,\n",
    "                                       class_mode='categorical',\n",
    "                                       target_size=(128,128))\n",
    "\n",
    "valid_gen = datagen.flow_from_dataframe(valid_df,\n",
    "                                       x_col='path',\n",
    "                                       y_col='label',\n",
    "                                       batch_size=32,\n",
    "                                       shuffle=True,\n",
    "                                       seed=2,\n",
    "                                       class_mode='categorical',\n",
    "                                       target_size=(128,128))\n",
    "\n",
    "test_gen = datagen.flow_from_dataframe(test_df,\n",
    "                                       x_col='path',\n",
    "                                       y_col='label',\n",
    "                                       batch_size=32,\n",
    "                                       shuffle=True,\n",
    "                                       seed=3,\n",
    "                                       class_mode='categorical',\n",
    "                                       target_size=(128,128))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "chkpnt = callbacks.ModelCheckpoint('checkpoint', save_weights_only=False, save_best_only=True, verbose=1)\n",
    "reduce_lr = callbacks.ReduceLROnPlateau(monitor='val_accuracy', factor=0.1, patience=4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = keras.Sequential()\n",
    "model.add(Conv2D(64, (3,3), activation='relu', input_shape=(128,128,3)))\n",
    "model.add(Conv2D(32, (3,3), activation='relu'))\n",
    "model.add(Conv2D(32, (3,3), activation='relu'))\n",
    "model.add(MaxPool2D((2,2)))\n",
    "\n",
    "model.add(Conv2D(64, (3,3), activation='relu'))\n",
    "model.add(Conv2D(32, (3,3), activation='relu'))\n",
    "model.add(Conv2D(32, (3,3), activation='relu'))\n",
    "model.add(MaxPool2D((2,2)))\n",
    "\n",
    "model.add(Flatten())\n",
    "model.add(Dense(1000, activation='relu'))\n",
    "model.add(Dense(100, activation='relu'))\n",
    "model.add(Dense(2, activation='softmax'))\n",
    "\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_6\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_32 (Conv2D)           (None, 126, 126, 64)      1792      \n",
      "_________________________________________________________________\n",
      "conv2d_33 (Conv2D)           (None, 124, 124, 32)      18464     \n",
      "_________________________________________________________________\n",
      "conv2d_34 (Conv2D)           (None, 122, 122, 32)      9248      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_13 (MaxPooling (None, 61, 61, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_35 (Conv2D)           (None, 59, 59, 64)        18496     \n",
      "_________________________________________________________________\n",
      "conv2d_36 (Conv2D)           (None, 57, 57, 32)        18464     \n",
      "_________________________________________________________________\n",
      "conv2d_37 (Conv2D)           (None, 55, 55, 32)        9248      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_14 (MaxPooling (None, 27, 27, 32)        0         \n",
      "_________________________________________________________________\n",
      "flatten_6 (Flatten)          (None, 23328)             0         \n",
      "_________________________________________________________________\n",
      "dense_18 (Dense)             (None, 1000)              23329000  \n",
      "_________________________________________________________________\n",
      "dense_19 (Dense)             (None, 100)               100100    \n",
      "_________________________________________________________________\n",
      "dense_20 (Dense)             (None, 2)                 202       \n",
      "=================================================================\n",
      "Total params: 23,505,014\n",
      "Trainable params: 23,505,014\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:sample_weight modes were coerced from\n",
      "  ...\n",
      "    to  \n",
      "  ['...']\n",
      "WARNING:tensorflow:sample_weight modes were coerced from\n",
      "  ...\n",
      "    to  \n",
      "  ['...']\n",
      "Train for 776 steps, validate for 44 steps\n",
      "Epoch 1/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 0.0843 - accuracy: 0.9712\n",
      "Epoch 00001: val_loss improved from inf to 0.14571, saving model to checkpoint\n",
      "INFO:tensorflow:Assets written to: checkpoint\\assets\n",
      "776/776 [==============================] - 213s 274ms/step - loss: 0.0843 - accuracy: 0.9712 - val_loss: 0.1457 - val_accuracy: 0.9470\n",
      "Epoch 2/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 0.0642 - accuracy: 0.9788\n",
      "Epoch 00002: val_loss did not improve from 0.14571\n",
      "776/776 [==============================] - 235s 302ms/step - loss: 0.0642 - accuracy: 0.9788 - val_loss: 0.2042 - val_accuracy: 0.9434\n",
      "Epoch 3/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 0.0378 - accuracy: 0.9878\n",
      "Epoch 00003: val_loss did not improve from 0.14571\n",
      "776/776 [==============================] - 222s 286ms/step - loss: 0.0378 - accuracy: 0.9878 - val_loss: 0.2020 - val_accuracy: 0.9514\n",
      "Epoch 4/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 0.0230 - accuracy: 0.9930\n",
      "Epoch 00004: val_loss did not improve from 0.14571\n",
      "776/776 [==============================] - 210s 270ms/step - loss: 0.0230 - accuracy: 0.9930 - val_loss: 0.2884 - val_accuracy: 0.9456\n",
      "Epoch 5/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 0.0183 - accuracy: 0.9944\n",
      "Epoch 00005: val_loss did not improve from 0.14571\n",
      "776/776 [==============================] - 249s 321ms/step - loss: 0.0183 - accuracy: 0.9944 - val_loss: 0.3998 - val_accuracy: 0.9427\n",
      "Epoch 6/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 0.0065 - accuracy: 0.9982\n",
      "Epoch 00006: val_loss did not improve from 0.14571\n",
      "776/776 [==============================] - 612s 789ms/step - loss: 0.0065 - accuracy: 0.9982 - val_loss: 0.4984 - val_accuracy: 0.9521\n",
      "Epoch 7/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 0.0143 - accuracy: 0.9963\n",
      "Epoch 00007: val_loss did not improve from 0.14571\n",
      "776/776 [==============================] - 302s 389ms/step - loss: 0.0143 - accuracy: 0.9963 - val_loss: 0.3333 - val_accuracy: 0.9427\n",
      "Epoch 8/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 0.0168 - accuracy: 0.9958\n",
      "Epoch 00008: val_loss did not improve from 0.14571\n",
      "776/776 [==============================] - 163s 210ms/step - loss: 0.0168 - accuracy: 0.9958 - val_loss: 0.4591 - val_accuracy: 0.9441\n",
      "Epoch 9/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 0.0117 - accuracy: 0.9972\n",
      "Epoch 00009: val_loss did not improve from 0.14571\n",
      "776/776 [==============================] - 185s 238ms/step - loss: 0.0117 - accuracy: 0.9972 - val_loss: 0.3670 - val_accuracy: 0.9499\n",
      "Epoch 10/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 0.0038 - accuracy: 0.9991\n",
      "Epoch 00010: val_loss did not improve from 0.14571\n",
      "776/776 [==============================] - 189s 244ms/step - loss: 0.0038 - accuracy: 0.9991 - val_loss: 0.5273 - val_accuracy: 0.9557\n",
      "Epoch 11/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 0.0105 - accuracy: 0.9973\n",
      "Epoch 00011: val_loss did not improve from 0.14571\n",
      "776/776 [==============================] - 189s 243ms/step - loss: 0.0105 - accuracy: 0.9973 - val_loss: 0.4354 - val_accuracy: 0.9485\n",
      "Epoch 12/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 0.0080 - accuracy: 0.9985\n",
      "Epoch 00012: val_loss did not improve from 0.14571\n",
      "776/776 [==============================] - 187s 241ms/step - loss: 0.0080 - accuracy: 0.9985 - val_loss: 0.3587 - val_accuracy: 0.9470\n",
      "Epoch 13/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 8.7758e-04 - accuracy: 0.9998\n",
      "Epoch 00013: val_loss did not improve from 0.14571\n",
      "776/776 [==============================] - 188s 243ms/step - loss: 8.7660e-04 - accuracy: 0.9998 - val_loss: 0.4657 - val_accuracy: 0.9528\n",
      "Epoch 14/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 0.0185 - accuracy: 0.9952\n",
      "Epoch 00014: val_loss did not improve from 0.14571\n",
      "776/776 [==============================] - 201s 259ms/step - loss: 0.0184 - accuracy: 0.9952 - val_loss: 0.3791 - val_accuracy: 0.9521\n",
      "Epoch 15/15\n",
      "775/776 [============================>.] - ETA: 0s - loss: 6.3266e-04 - accuracy: 0.9999\n",
      "Epoch 00015: val_loss did not improve from 0.14571\n",
      "776/776 [==============================] - 233s 300ms/step - loss: 6.3187e-04 - accuracy: 0.9999 - val_loss: 0.4182 - val_accuracy: 0.9507\n"
     ]
    }
   ],
   "source": [
    "h = model.fit(train_gen, epochs=15, verbose=1, callbacks=[chkpnt, reduce_lr], validation_data=valid_gen)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.save('final_model.hdf5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x210860e9a20>]"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(h.history['val_accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:sample_weight modes were coerced from\n",
      "  ...\n",
      "    to  \n",
      "  ['...']\n",
      "43/43 [==============================] - 2s 39ms/step - loss: 0.7180 - accuracy: 0.9273\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.7179992913412249, 0.9273256]"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(test_gen)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.models import load_model\n",
    "m2 = load_model('checkpoint')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:sample_weight modes were coerced from\n",
      "  ...\n",
      "    to  \n",
      "  ['...']\n",
      "43/43 [==============================] - 2s 46ms/step - loss: 0.2338 - accuracy: 0.9201\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.23376682201443716, 0.92005813]"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "m2.evaluate(test_gen)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}