002_Exploratory_Data_Analysis.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Exploratory Data Analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Libraries to Import"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "import os\n",
    "import gc\n",
    "\n",
    "from PIL import Image\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from keras import layers\n",
    "from keras.preprocessing import image"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Set filepaths for train/test/validation images."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_dir = '/Users/arencarpenter/Desktop/Detecting_ALL_with_CNN/Data/training'\n",
    "test_dir = '/Users/arencarpenter/Desktop/Detecting_ALL_with_CNN/Data/testing'\n",
    "val_dir = '/Users/arencarpenter/Desktop/Detecting_ALL_with_CNN/Data/validation'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create lists for normal and ALL images pulled from the respective directories."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "normal_imgs = [fn for fn in os.listdir(f'{train_dir}/normal')]\n",
    "all_imgs = [fn for fn in os.listdir(f'{train_dir}/all')]\n",
    "\n",
    "select_normal = np.random.choice(normal_imgs, 3, replace = False)\n",
    "select_all = np.random.choice(all_imgs, 3, replace = False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Plot random images from each class."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 6 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize = (8,6))\n",
    "\n",
    "for i in range(6):\n",
    "    if i < 3:\n",
    "        fp = f'{train_dir}/normal/{select_normal[i]}'\n",
    "        label = 'Normal'\n",
    "    else:\n",
    "        fp = f'{train_dir}/all/{select_all[i-3]}'\n",
    "        label = 'ALL'\n",
    "    ax = fig.add_subplot(2, 3, i+1)\n",
    "    fn = image.load_img(fp, target_size = (100,100),\n",
    "                        color_mode='rgb')\n",
    "    plt.imshow(fn, cmap='Greys_r')\n",
    "    plt.title(label)\n",
    "    plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Use length of class lists to plot original dataset class imbalance."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "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.bar(['Normal', 'ALL'], [len(normal_imgs), len(all_imgs)])\n",
    "plt.title('Original Class Imbalance')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Visualize downsampled class imbalance. Final split was 0.57/0.43."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "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.bar(['Normal', 'ALL'], [1642, 2156])\n",
    "plt.title('Downsampled Class Imbalanced')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create function to process images to NumPy arrays to create Mean Images."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def img2np(pwd, fn_list, size = (32, 32)):\n",
    "    for fn in fn_list:\n",
    "        fp = pwd + fn\n",
    "        current_image = image.load_img(fp, \n",
    "                                       target_size = size, \n",
    "                                       color_mode = 'grayscale')\n",
    "        img_ts = image.img_to_array(current_image)\n",
    "        img_ts = [img_ts.ravel()]\n",
    "        try:\n",
    "            full_mat = np.concatenate((full_mat, img_ts))\n",
    "        except UnboundLocalError: \n",
    "            full_mat = img_ts\n",
    "    return full_mat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "normal_images = img2np(f'{train_dir}/normal/', normal_imgs)\n",
    "all_images = img2np(f'{train_dir}/all/', all_imgs)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Just select 25 images to make comparison cleaner."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "normal_images = normal_images[:25]\n",
    "all_images = all_images[:25]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create function to plot Mean Images."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def find_mean_img(image_vecs, title, size = (32,32)):\n",
    "    mean_img = np.mean(image_vecs, axis = 0)\n",
    "    mean_img = mean_img.reshape(size)\n",
    "    plt.imshow(mean_img, vmin=0, vmax=255)\n",
    "    plt.title(f'Average {title}')\n",
    "    plt.axis('off')\n",
    "    plt.show()\n",
    "    return mean_img"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "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": [
    "norm_mean = find_mean_img(normal_images, 'Normal')\n",
    "pneu_mean = find_mean_img(all_images, 'ALL')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The diversity of cell interiors is such that mean images can't detect differences between the classes, however we can clearly see that ALL cells are much larger than normal cells. This makes sense as ALL cells have unregulated cell growth and would thus be larger."
   ]
  }
 ],
 "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}