data_preparation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Preprocessing\n",
    "# *Weakly Supervised Learning for Industrial Optical Inspection*\n",
    "\n",
    "Data could be found at http://resources.mpi-inf.mpg.de/conferences/dagm/2007/prizes.html, to download the 12 zip-files we type\n",
    "```\n",
    "wget http://resources.mpi-inf.mpg.de/conferences/dagm/2007/Class{1..6}.zip\n",
    "wget http://resources.mpi-inf.mpg.de/conferences/dagm/2007/Class{1..6}_def.zip\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# author: Zilong Zhao"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import glob\n",
    "import seaborn as sns\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "from keras.preprocessing.image import ImageDataGenerator,array_to_img, img_to_array, load_img\n",
    "import math\n",
    "from tqdm import tqdm\n",
    "import random\n",
    "import shutil"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "raw_data_dir = '/opt/Daten/DAGM/Raw_data/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "def check_barplot(data_dir):\n",
    "    class1=[]\n",
    "    class2=[]\n",
    "    class3=[]\n",
    "    class4=[]\n",
    "    class5=[]\n",
    "    class6=[]\n",
    "    class1_def=[]\n",
    "    class2_def=[]\n",
    "    class3_def=[]\n",
    "    class4_def=[]\n",
    "    class5_def=[]\n",
    "    class6_def=[]\n",
    "\n",
    "    for filename in glob.iglob(data_dir + '**/*.*', recursive=True):\n",
    "        if 'Class1/' in filename:\n",
    "            class1.append(filename)\n",
    "        elif 'Class2/' in filename:\n",
    "            class2.append(filename)\n",
    "        elif 'Class3/' in filename:\n",
    "            class3.append(filename)\n",
    "        elif 'Class4/' in filename:\n",
    "            class4.append(filename)\n",
    "        elif 'Class5/' in filename:\n",
    "            class5.append(filename)\n",
    "        elif 'Class6/' in filename:\n",
    "            class6.append(filename)\n",
    "        elif 'Class1_def/' in filename:\n",
    "            class1_def.append(filename)\n",
    "        elif 'Class2_def/' in filename:\n",
    "            class2_def.append(filename)\n",
    "        elif 'Class3_def/' in filename:\n",
    "            class3_def.append(filename)\n",
    "        elif 'Class4_def/' in filename:\n",
    "            class4_def.append(filename)\n",
    "        elif 'Class5_def/' in filename:\n",
    "            class5_def.append(filename)\n",
    "        else:\n",
    "            class6_def.append(filename)\n",
    "    class_name=sorted(['Class' + str(n) for n in range(1, 7)] + ['Class' + str(n) + '_def' for n in range(1,7)])\n",
    "    class_count=[len(class1), len(class1_def), \n",
    "                 len(class2), len(class2_def),\n",
    "                 len(class3), len(class3_def),\n",
    "                 len(class4), len(class4_def),\n",
    "                 len(class5), len(class5_def),\n",
    "                 len(class6), len(class6_def)]\n",
    "    plt.figure(figsize = (12,6))\n",
    "    sns.barplot(class_name, class_count, alpha = 0.9)\n",
    "    plt.xticks(rotation = 'vertical')\n",
    "    plt.xlabel('Class name', fontsize =12)\n",
    "    plt.ylabel('Counts', fontsize = 12)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Let's check how many pics we have for each class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "check_barplot(data_dir=raw_data_dir)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The barplot shows an extream unbalance in the classes with defectness. To prevent overfitting, we need to do some data augmentation job."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data augmentation\n",
    "Using Keras, we can easily rotate, flip, shift the image with\n",
    "```\n",
    "keras.preprocessing.image.ImageDataGenerator()\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "  0%|          | 0/12 [00:00<?, ?it/s]\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "  8%|▊         | 1/12 [00:08<01:32,  8.45s/it]\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      " 17%|█▋        | 2/12 [00:53<03:14, 19.42s/it]\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      " 25%|██▌       | 3/12 [01:01<02:24, 16.07s/it]\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      " 33%|███▎      | 4/12 [01:45<03:16, 24.52s/it]\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      " 42%|████▏     | 5/12 [01:54<02:17, 19.60s/it]\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      " 50%|█████     | 6/12 [02:37<02:39, 26.61s/it]\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      " 58%|█████▊    | 7/12 [02:45<01:45, 21.03s/it]\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      " 67%|██████▋   | 8/12 [03:27<01:49, 27.45s/it]\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      " 75%|███████▌  | 9/12 [03:35<01:04, 21.64s/it]\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      " 83%|████████▎ | 10/12 [04:18<00:56, 28.05s/it]\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      " 92%|█████████▏| 11/12 [04:27<00:22, 22.24s/it]\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "100%|██████████| 12/12 [05:12<00:00, 29.05s/it]\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A\u001b[A"
     ]
    }
   ],
   "source": [
    "datagen = ImageDataGenerator(\n",
    "        rotation_range=30,\n",
    "        width_shift_range=0.1,\n",
    "        height_shift_range=0.1,\n",
    "        shear_range=0.1,\n",
    "        horizontal_flip=True,\n",
    "        vertical_flip=True,\n",
    "        fill_mode='nearest')\n",
    "\n",
    "\n",
    "target_class_count = 1200\n",
    "\n",
    "dest_train_dir='/opt/Daten/DAGM/train_data/'\n",
    "\n",
    "def tqdm_enumerate(iter):\n",
    "    i = 0\n",
    "    for y in tqdm(iter):\n",
    "        yield i, y\n",
    "        i += 1\n",
    "\n",
    "for j, c in tqdm_enumerate(class_count):\n",
    "    dest_lab_dir=os.path.join(dest_train_dir, class_name[j])\n",
    "    src_lab_dir=os.path.join(raw_data_dir, class_name[j])\n",
    "    \n",
    "    \n",
    "    if not os.path.exists(dest_lab_dir):\n",
    "        os.makedirs(dest_lab_dir)\n",
    "    random_file=random.choices(population=[x for x in os.listdir(src_lab_dir) if '.png' in x], k=target_class_count-c)\n",
    "    for file in random_file:\n",
    "        img=load_img(os.path.join(src_lab_dir,file))\n",
    "        x=img_to_array(img)\n",
    "        x=x.reshape((1,) + x.shape)\n",
    "        for batch in datagen.flow(x, batch_size=1, save_to_dir=dest_lab_dir, save_format='jpg', save_prefix=file):\n",
    "            if target_class_count - len(os.listdir(dest_lab_dir)) > 0:\n",
    "                break\n",
    "                    \n",
    "            "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [],
   "source": [
    "!scp -r /opt/Daten/DAGM/Raw_data/. /opt/Daten/DAGM/train_data/."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [],
   "source": [
    "dest_test_dir='/opt/Daten/DAGM/test_data/'\n",
    "\n",
    "for d in os.listdir(dest_train_dir):\n",
    "    files=os.listdir(os.path.join(dest_train_dir, d))\n",
    "    dest_lab_dir=os.path.join(dest_test_dir, d)\n",
    "    if not os.path.exists(dest_lab_dir):\n",
    "        os.makedirs(dest_lab_dir)\n",
    "    random_file=random.sample(population=[x for x in files], k=300)\n",
    "    for rf in random_file:\n",
    "        shutil.move(dest_train_dir+d+'/'+rf, dest_test_dir+d+'/'+rf)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Now let's check again"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# the training data\n",
    "check_barplot(data_dir='/opt/Daten/DAGM/train_data/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# the test data\n",
    "check_barplot(data_dir='/opt/Daten/DAGM/test_data/')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "After augmenting the data, we got balanced training- and test datasets. Next step, we will use the **retrain.py** script, which is officially provided by TensorFlow and could be downloaded with\n",
    "```\n",
    "wget https://github.com/tensorflow/hub/raw/master/examples/image_retraining/retrain.py\n",
    "```\n",
    "\n",
    "To start it, simply type\n",
    "```\n",
    "python retrain.py --image_dir /opt/Daten/DAGM/train_data\n",
    "```\n",
    "As a default setting, it will download the **Inception V3** model from [TensorFlow Hub](https://tfhub.dev/) and train with our pre-processed training data automatically.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda root]",
   "language": "python",
   "name": "conda-root-py"
  },
  "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
}