{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import glob\n",
    "import numpy as np\n",
    "import random\n",
    "\n",
    "import cv2\n",
    "import matplotlib.pyplot as plt\n",
    "from PIL import Image\n",
    "\n",
    "from torchvision import transforms as pil_transforms\n",
    "from torchvision.transforms import functional as F_pil\n",
    "\n",
    "import sys\n",
    "sys.path.insert(0, '..')\n",
    "from opencv_transforms import transforms\n",
    "from opencv_transforms import functional as F\n",
    "\n",
    "from setup_testing_directory import get_testing_directory"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "datadir = get_testing_directory()\n",
    "print(datadir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_images = glob.glob(datadir + '/**/*.JPEG', recursive=True)\n",
    "train_images.sort()\n",
    "print('Number of training images: {:,}'.format(len(train_images)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "random.seed(1)\n",
    "imfile = random.choice(train_images)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_pil_and_opencv(pil_image, opencv_image, orientation='row'):\n",
    "    if orientation == 'row':\n",
    "        rows, cols = 1,3\n",
    "        size = (8, 4)\n",
    "    else: \n",
    "        rows, cols = 3,1\n",
    "        size = (12, 6)\n",
    "    fig, axes = plt.subplots(rows, cols,figsize=size)\n",
    "    ax = axes[0]\n",
    "    ax.imshow(pil_image)\n",
    "    ax.set_title('PIL')\n",
    "\n",
    "    ax = axes[1]\n",
    "    ax.imshow(opencv_image)\n",
    "    ax.set_title('opencv')\n",
    "\n",
    "    ax = axes[2]\n",
    "    l1 = np.abs(pil_image - opencv_image).mean(axis=2)\n",
    "    ax.imshow(l1)\n",
    "    ax.set_title('| PIL - opencv|\\nMAE:{:.4f}'.format(l1.mean()))\n",
    "    plt.tight_layout()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pil_image = Image.open(imfile)\n",
    "image = cv2.cvtColor(cv2.imread(imfile, 1), cv2.COLOR_BGR2RGB)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plot_pil_and_opencv(pil_image, image)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pil_resized = pil_transforms.Resize((224, 224))(pil_image)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "resized = transforms.Resize(224)(image)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plot_pil_and_opencv(pil_resized, resized)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def L1(pil: Image, image: np.ndarray) -> float:\n",
    "    return np.mean(np.abs(np.asarray(pil) - image))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "TOL = 1e-4\n",
    "\n",
    "l1 = L1(pil_resized, resized)\n",
    "assert l1 - 88.9559 < TOL"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "random.seed(1)\n",
    "pil = pil_transforms.RandomRotation(10)(pil_image)\n",
    "random.seed(1)\n",
    "np_img = transforms.RandomRotation(10)(image)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plot_pil_and_opencv(pil, np_img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pil = pil_transforms.FiveCrop((224, 224))(pil_image)\n",
    "cv  = transforms.FiveCrop((224,224))(image)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pil_stacked = np.hstack([np.asarray(i) for i in pil])\n",
    "cv_stacked = np.hstack(cv)\n",
    "\n",
    "plot_pil_and_opencv(pil_stacked, cv_stacked, orientation='col')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pil_stacked.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "l1"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "opencv_transforms",
   "language": "python",
   "name": "opencv_transforms"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}