cnn_argumented.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "source": [
    "# Argument the image to solve overfitting\n",
    "run 'cnn_origin.ipynb' to make subdirectory and split data\n"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "outputs": [],
   "source": [
    "import os\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "import matplotlib.pyplot as plt"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "outputs": [],
   "source": [
    "source_path = './data/cats-and-dogs/PetImages/'\n",
    "source_path_dogs = os.path.join(source_path, 'Dog')\n",
    "source_path_cats = os.path.join(source_path, 'Cat')\n",
    "root_dir = './data'\n",
    "training_path = os.path.join(root_dir, \"training\")\n",
    "test_path = os.path.join(root_dir, \"validation\")\n",
    "training_path_cats = os.path.join(training_path, \"cats\")\n",
    "training_path_dogs = os.path.join(training_path, \"dogs\")\n",
    "test_path_cats = os.path.join(test_path, \"cats\")\n",
    "test_path_dogs = os.path.join(test_path, \"dogs\")"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "From the previous model, the plot shows that the model is overfitting (validation acc is much lower then training acc), the following added parameter will reduce this situation (notice that different overfiting situation has different technique to utilize)"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 22498 images belonging to 2 classes.\n",
      "Found 2500 images belonging to 2 classes.\n"
     ]
    }
   ],
   "source": [
    "train_datagen = ImageDataGenerator(rescale=1./255,\n",
    "                    rotation_range=40,\n",
    "                    width_shift_range=0.2,\n",
    "                    height_shift_range=0.2,\n",
    "                    shear_range=0.2,\n",
    "                    zoom_range=0.2,\n",
    "                    horizontal_flip=True,\n",
    "                    fill_mode='nearest')\n",
    "train_generator = train_datagen.flow_from_directory(directory=training_path,\n",
    "                              batch_size=100,\n",
    "                              class_mode='binary',\n",
    "                              target_size=(150, 150))\n",
    "validation_datagen = ImageDataGenerator(rescale=1./255)\n",
    "validation_generator = validation_datagen.flow_from_directory(directory=test_path,\n",
    "                              batch_size=100,\n",
    "                              class_mode='binary',\n",
    "                              target_size=(150, 150))"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "set the same model to see the difference:"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/25\n",
      "225/225 [==============================] - 111s 490ms/step - loss: 0.6739 - accuracy: 0.6079 - val_loss: 0.5464 - val_accuracy: 0.7208\n",
      "Epoch 2/25\n",
      "225/225 [==============================] - 111s 493ms/step - loss: 0.6003 - accuracy: 0.6716 - val_loss: 0.5234 - val_accuracy: 0.7372\n",
      "Epoch 3/25\n",
      "225/225 [==============================] - 112s 498ms/step - loss: 0.5708 - accuracy: 0.7018 - val_loss: 0.5060 - val_accuracy: 0.7568\n",
      "Epoch 4/25\n",
      "225/225 [==============================] - 113s 504ms/step - loss: 0.5466 - accuracy: 0.7209 - val_loss: 0.4556 - val_accuracy: 0.7932\n",
      "Epoch 5/25\n",
      "225/225 [==============================] - 115s 510ms/step - loss: 0.5313 - accuracy: 0.7370 - val_loss: 0.4744 - val_accuracy: 0.7636\n",
      "Epoch 6/25\n",
      "225/225 [==============================] - 114s 508ms/step - loss: 0.5189 - accuracy: 0.7428 - val_loss: 0.4339 - val_accuracy: 0.7956\n",
      "Epoch 7/25\n",
      "225/225 [==============================] - 114s 506ms/step - loss: 0.5059 - accuracy: 0.7510 - val_loss: 0.4205 - val_accuracy: 0.8048\n",
      "Epoch 8/25\n",
      "225/225 [==============================] - 114s 505ms/step - loss: 0.4902 - accuracy: 0.7650 - val_loss: 0.4657 - val_accuracy: 0.7860\n",
      "Epoch 9/25\n",
      "225/225 [==============================] - 114s 505ms/step - loss: 0.4809 - accuracy: 0.7692 - val_loss: 0.4092 - val_accuracy: 0.8052\n",
      "Epoch 10/25\n",
      "225/225 [==============================] - 114s 506ms/step - loss: 0.4728 - accuracy: 0.7740 - val_loss: 0.3759 - val_accuracy: 0.8348\n",
      "Epoch 11/25\n",
      "225/225 [==============================] - 114s 506ms/step - loss: 0.4564 - accuracy: 0.7826 - val_loss: 0.4283 - val_accuracy: 0.8068\n",
      "Epoch 12/25\n",
      "225/225 [==============================] - 114s 508ms/step - loss: 0.4482 - accuracy: 0.7866 - val_loss: 0.3534 - val_accuracy: 0.8440\n",
      "Epoch 13/25\n",
      "225/225 [==============================] - 115s 509ms/step - loss: 0.4402 - accuracy: 0.7948 - val_loss: 0.3542 - val_accuracy: 0.8424\n",
      "Epoch 14/25\n",
      "225/225 [==============================] - 115s 510ms/step - loss: 0.4386 - accuracy: 0.7982 - val_loss: 0.3418 - val_accuracy: 0.8600\n",
      "Epoch 15/25\n",
      "225/225 [==============================] - 115s 512ms/step - loss: 0.4196 - accuracy: 0.8046 - val_loss: 0.3452 - val_accuracy: 0.8428\n",
      "Epoch 16/25\n",
      "225/225 [==============================] - 115s 509ms/step - loss: 0.4140 - accuracy: 0.8123 - val_loss: 0.3098 - val_accuracy: 0.8632\n",
      "Epoch 17/25\n",
      "225/225 [==============================] - 115s 510ms/step - loss: 0.4072 - accuracy: 0.8139 - val_loss: 0.3431 - val_accuracy: 0.8524\n",
      "Epoch 18/25\n",
      "225/225 [==============================] - 115s 509ms/step - loss: 0.4013 - accuracy: 0.8187 - val_loss: 0.3414 - val_accuracy: 0.8476\n",
      "Epoch 19/25\n",
      "225/225 [==============================] - 114s 507ms/step - loss: 0.3925 - accuracy: 0.8225 - val_loss: 0.3067 - val_accuracy: 0.8684\n",
      "Epoch 20/25\n",
      "225/225 [==============================] - 115s 510ms/step - loss: 0.3887 - accuracy: 0.8272 - val_loss: 0.2831 - val_accuracy: 0.8816\n",
      "Epoch 21/25\n",
      "225/225 [==============================] - 116s 514ms/step - loss: 0.3852 - accuracy: 0.8299 - val_loss: 0.3131 - val_accuracy: 0.8692\n",
      "Epoch 22/25\n",
      "225/225 [==============================] - 116s 516ms/step - loss: 0.3784 - accuracy: 0.8300 - val_loss: 0.3643 - val_accuracy: 0.8624\n",
      "Epoch 23/25\n",
      "225/225 [==============================] - 115s 512ms/step - loss: 0.3784 - accuracy: 0.8325 - val_loss: 0.3167 - val_accuracy: 0.8656\n",
      "Epoch 24/25\n",
      "225/225 [==============================] - 115s 510ms/step - loss: 0.3665 - accuracy: 0.8394 - val_loss: 0.3715 - val_accuracy: 0.8396\n",
      "Epoch 25/25\n",
      "225/225 [==============================] - 116s 516ms/step - loss: 0.3691 - accuracy: 0.8369 - val_loss: 0.2983 - val_accuracy: 0.8784\n"
     ]
    }
   ],
   "source": [
    "model = tf.keras.models.Sequential([\n",
    "    tf.keras.layers.Conv2D(16, (3, 3), activation='relu', input_shape=(150, 150, 3)),\n",
    "    tf.keras.layers.MaxPooling2D(2, 2),\n",
    "    tf.keras.layers.Conv2D(32, (3, 3), activation='relu'),\n",
    "    tf.keras.layers.MaxPooling2D(2, 2),\n",
    "    tf.keras.layers.Conv2D(64, (3, 3), activation='relu'),\n",
    "    tf.keras.layers.MaxPooling2D(2, 2),\n",
    "    tf.keras.layers.Flatten(),\n",
    "    tf.keras.layers.Dense(512, activation='relu'),\n",
    "    tf.keras.layers.Dense(1, activation='sigmoid')\n",
    "])\n",
    "model.compile(optimizer=tf.keras.optimizers.RMSprop(learning_rate=0.001),\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "history = model.fit(train_generator,\n",
    "                    epochs=25,\n",
    "                    verbose=1,\n",
    "                    validation_data=validation_generator)"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "acc = history.history['accuracy']\n",
    "val_acc = history.history['val_accuracy']\n",
    "loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "epochs = range(len(acc))\n",
    "plt.plot(epochs, acc, 'r', \"Training Accuracy\")\n",
    "plt.plot(epochs, val_acc, 'b', \"Validation Accuracy\")\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.show()\n",
    "print(\"\")\n",
    "\n",
    "plt.plot(epochs, loss, 'r', \"Training Loss\")\n",
    "plt.plot(epochs, val_loss, 'b', \"Validation Loss\")\n",
    "plt.show()\n"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "The validation acc is close to the training acc, but the accuracy still surrounds with 80%, this due to training scale, iteration times, and also some engineering. Next, we will focus on transferring learning"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "outputs": [],
   "source": [
    "model.save_weights('./checkpoints/cnn_argument_data')"
   ],
   "metadata": {
    "collapsed": false
   }
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}