prepare-data.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ubuntu/anaconda3/envs/tensorflow_p36/lib/python3.6/importlib/_bootstrap.py:219: RuntimeWarning: compiletime version 3.5 of module 'tensorflow.python.framework.fast_tensor_util' does not match runtime version 3.6\n",
      "  return f(*args, **kwds)\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import pickle\n",
    "import cv2\n",
    "import numpy as np\n",
    "from mtcnn.mtcnn import MTCNN\n",
    "from matplotlib import pyplot as plt\n",
    "\n",
    "detector = MTCNN()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "SMILE_FOLDER = '../datasets/genki4k/'\n",
    "F_SMILE_FOLDER = './data/smile_data/'\n",
    "NUM_SMILE_IMAGE = 4000\n",
    "SMILE_SIZE = 48\n",
    "\n",
    "X = []\n",
    "with open(SMILE_FOLDER + \"labels.txt\") as f:\n",
    "    for i in range(NUM_SMILE_IMAGE):\n",
    "        fileName = SMILE_FOLDER + \"files/file\" + str(\"%04d\" % (1+i,)) + \".jpg\"\n",
    "#             print(fileName)\n",
    "        img = cv2.imread(fileName)\n",
    "        result = detector.detect_faces(img)\n",
    "        if not result:\n",
    "            continue\n",
    "        face_position = result[0].get('box')\n",
    "        x = face_position[0]\n",
    "        y = face_position[1]\n",
    "        w = face_position[2]\n",
    "        h = face_position[3]\n",
    "        img = img[y:y+h, x:x+w]\n",
    "        if(img.size==0):\n",
    "            continue;\n",
    "        img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)\n",
    "        img = cv2.resize(img, (48, 48))\n",
    "        T = np.zeros([SMILE_SIZE, SMILE_SIZE, 1])\n",
    "        T[:, :, 0] = img\n",
    "        l = f.readline()\n",
    "        label = (int)(l.split()[0])\n",
    "        X.append((T, label))\n",
    "for _ in range(10):\n",
    "    np.random.shuffle(X)\n",
    "\n",
    "train_data, test_data = X[:3000], X[3000:]\n",
    "\n",
    "np.save(F_SMILE_FOLDER + 'train.npy', train_data)\n",
    "np.save(F_SMILE_FOLDER + 'test.npy', test_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "i: 0\n",
      "i: 10000\n",
      "i: 20000\n",
      "i: 30000\n",
      "i: 40000\n",
      "i: 50000\n",
      "i: 60000\n",
      "i: 70000\n",
      "i: 80000\n",
      "i: 90000\n",
      "i: 100000\n",
      "i: 110000\n",
      "i: 120000\n",
      "i: 130000\n",
      "i: 140000\n",
      "i: 150000\n",
      "i: 160000\n"
     ]
    }
   ],
   "source": [
    "F_AGE_FOLDER = './data/age_data/'\n",
    "F_GENDER_FOLDER = './data/gender_data/'\n",
    "\n",
    "f = open(os.path.expanduser(\"~\")+'/coding/cnn/multi-task-learning/pkl_folder/imdb-age-gender.pkl', 'rb')\n",
    "imdbAgeGender = pickle.load(f)\n",
    "age = imdbAgeGender['age']\n",
    "gender = imdbAgeGender['gender']\n",
    "imgs = imdbAgeGender['imgs']\n",
    "f.close()\n",
    "n = len(gender)\n",
    "\n",
    "X_age = []\n",
    "X_gender = []\n",
    "\n",
    "for i in range(n):\n",
    "    if i%10000==0:\n",
    "        print(\"i: {}\".format(i))\n",
    "#         fileName = IMDB_FOLDER + str(i) + '.jpg'\n",
    "    img = cv2.imread(imgs[i])\n",
    "    result = detector.detect_faces(img)\n",
    "    if not result:\n",
    "        continue\n",
    "    face_position = result[0].get('box')\n",
    "    x = face_position[0]\n",
    "    y = face_position[1]\n",
    "    w = face_position[2]\n",
    "    h = face_position[3]\n",
    "    img = img[y:y+h, x:x+w]\n",
    "    if(img.size==0):\n",
    "        continue;\n",
    "    img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)\n",
    "    img = cv2.resize(img,(48, 48))\n",
    "\n",
    "    T = np.zeros([48, 48, 1])\n",
    "    T[:, :, 0] = img\n",
    "\n",
    "    label_age = age[i]\n",
    "    label_gender = gender[i]\n",
    "    X_age.append((T, label_age))\n",
    "    X_gender.append((T, label_gender))\n",
    "\n",
    "for _ in range(10):\n",
    "    np.random.shuffle(X_age)\n",
    "    np.random.shuffle(X_gender)\n",
    "\n",
    "train_data_age, test_data_age = X_age[:90000], X_age[90000:]\n",
    "train_data_gender, test_data_gender = X_gender[:90000], X_gender[90000:]\n",
    "\n",
    "np.save(F_AGE_FOLDER + 'train.npy', train_data_age)\n",
    "np.save(F_AGE_FOLDER + 'data.npy', X_age)\n",
    "np.save(F_AGE_FOLDER + 'test.npy', test_data_age)\n",
    "\n",
    "np.save(F_GENDER_FOLDER + 'train.npy', train_data_gender)\n",
    "np.save(F_GENDER_FOLDER + 'data.npy', X_gender)\n",
    "np.save(F_GENDER_FOLDER + 'test.npy', test_data_gender)"
   ]
  },
  {
   "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}