CarND-Traffic-Sign-Classifier-resubmit.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Self-Driving Car Engineer Nanodegree\n",
    "\n",
    "## Deep Learning\n",
    "\n",
    "## Project: Build a Traffic Sign Recognition Classifier\n",
    "\n",
    "In this notebook, a template is provided for you to implement your functionality in stages, which is required to successfully complete this project. If additional code is required that cannot be included in the notebook, be sure that the Python code is successfully imported and included in your submission if necessary. \n",
    "\n",
    "> **Note**: Once you have completed all of the code implementations, you need to finalize your work by exporting the iPython Notebook as an HTML document. Before exporting the notebook to html, all of the code cells need to have been run so that reviewers can see the final implementation and output. You can then export the notebook by using the menu above and navigating to  \\n\",\n",
    "    \"**File -> Download as -> HTML (.html)**. Include the finished document along with this notebook as your submission. \n",
    "\n",
    "In addition to implementing code, there is a writeup to complete. The writeup should be completed in a separate file, which can be either a markdown file or a pdf document. There is a [write up template](https://github.com/udacity/CarND-Traffic-Sign-Classifier-Project/blob/master/writeup_template.md) that can be used to guide the writing process. Completing the code template and writeup template will cover all of the [rubric points](https://review.udacity.com/#!/rubrics/481/view) for this project.\n",
    "\n",
    "The [rubric](https://review.udacity.com/#!/rubrics/481/view) contains \"Stand Out Suggestions\" for enhancing the project beyond the minimum requirements. The stand out suggestions are optional. If you decide to pursue the \"stand out suggestions\", you can include the code in this Ipython notebook and also discuss the results in the writeup file.\n",
    "\n",
    "\n",
    ">**Note:** Code and Markdown cells can be executed using the **Shift + Enter** keyboard shortcut. In addition, Markdown cells can be edited by typically double-clicking the cell to enter edit mode."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 0: Load The Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Images :  (34799, 32, 32, 3)\n"
     ]
    }
   ],
   "source": [
    "# Load pickled data\n",
    "import pickle\n",
    "\n",
    "# TODO: Fill this in based on where you saved the training and testing data\n",
    "# Below are the details of the training, validation and test data\n",
    "# main_folder is used to reduce typing time!\n",
    "\n",
    "training_file = \"./traffic-signs-data/train.p\"\n",
    "validation_file= \"./traffic-signs-data/valid.p\"\n",
    "testing_file = \"./traffic-signs-data/test.p\"\n",
    "\n",
    "with open(training_file, mode='rb') as f:\n",
    "    train = pickle.load(f)\n",
    "with open(validation_file, mode='rb') as f:\n",
    "    valid = pickle.load(f)\n",
    "with open(testing_file, mode='rb') as f:\n",
    "    test = pickle.load(f)\n",
    "    \n",
    "X_train, y_train = train['features'], train['labels']\n",
    "X_test, y_test = test['features'], test['labels']\n",
    "X_valid, y_valid = valid['features'], valid['labels']\n",
    "print('Images : ', X_train.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "\n",
    "## Step 1: Dataset Summary & Exploration\n",
    "\n",
    "The pickled data is a dictionary with 4 key/value pairs:\n",
    "\n",
    "- `'features'` is a 4D array containing raw pixel data of the traffic sign images, (num examples, width, height, channels).\n",
    "- `'labels'` is a 1D array containing the label/class id of the traffic sign. The file `signnames.csv` contains id -> name mappings for each id.\n",
    "- `'sizes'` is a list containing tuples, (width, height) representing the original width and height the image.\n",
    "- `'coords'` is a list containing tuples, (x1, y1, x2, y2) representing coordinates of a bounding box around the sign in the image. **THESE COORDINATES ASSUME THE ORIGINAL IMAGE. THE PICKLED DATA CONTAINS RESIZED VERSIONS (32 by 32) OF THESE IMAGES**\n",
    "\n",
    "Complete the basic data summary below. Use python, numpy and/or pandas methods to calculate the data summary rather than hard coding the results. For example, the [pandas shape method](http://pandas.pydata.org/pandas-docs/stable/generated/pandas.DataFrame.shape.html) might be useful for calculating some of the summary results. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Provide a Basic Summary of the Data Set Using Python, Numpy and/or Pandas"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training shape = (32, 32, 3)\n",
      "Number of training examples = 34799\n",
      "Number of validation examples = 4410\n",
      "Number of testing examples = 12630\n",
      "Image data shape = (32, 32, 3)\n",
      "Number of classes = 43\n"
     ]
    }
   ],
   "source": [
    "### Replace each question mark with the appropriate value. \n",
    "### Use python, pandas or numpy methods rather than hard coding the results\n",
    "import numpy as np\n",
    "import cv2\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# TODO: Number of training examples\n",
    "n_train = len(X_train)\n",
    "\n",
    "# TODO: Number of validation examples\n",
    "n_validation = len(X_valid)\n",
    "\n",
    "# TODO: Number of testing examples.\n",
    "n_test = len(X_test)\n",
    "\n",
    "# TODO: What's the shape of an traffic sign image?\n",
    "image_shape = X_train.shape[1:4]\n",
    "print(\"Training shape =\", image_shape)\n",
    "\n",
    "# TODO: How many unique classes/labels there are in the dataset.\n",
    "#n_classes = np.unique(y_train)[-1] + 1 # different way of getting the # classes\n",
    "n_classes = len(np.unique(y_train))\n",
    "\n",
    "print(\"Number of training examples =\", n_train)\n",
    "print(\"Number of validation examples =\", n_validation)\n",
    "print(\"Number of testing examples =\", n_test)\n",
    "print(\"Image data shape =\", image_shape)\n",
    "print(\"Number of classes =\", n_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75e6bd9160>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "### Data exploration visualization goes here.\n",
    "### Feel free to use as many code cells as needed.\n",
    "import random\n",
    "# Visualizations will be shown in the notebook.\n",
    "%matplotlib inline\n",
    "\n",
    "# show image of 10 random data points\n",
    "fig, axs = plt.subplots(1,5, figsize=(15, 6))\n",
    "fig.subplots_adjust(hspace = .2, wspace=.001)\n",
    "axs = axs.ravel()\n",
    "for i in range(5):\n",
    "    index = random.randint(0, len(X_train))\n",
    "    image = X_train[index]\n",
    "    axs[i].axis('off')\n",
    "    axs[i].imshow(image)\n",
    "    axs[i].set_title(y_train[index])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Include an exploratory visualization of the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "shape in/out: (32, 32, 3) (32, 32, 1, 3)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75ad6ae470>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def random_translate(img):\n",
    "    rows,cols,_ = img.shape\n",
    "    \n",
    "    # x and y directional translation\n",
    "    px = 2\n",
    "    dx,dy = np.random.randint(-px,px,2)\n",
    "    M = np.float32([[1,0,dx],[0,1,dy]])\n",
    "    dst = cv2.warpAffine(img,M,(cols,rows))\n",
    "    dst = dst[:,:,np.newaxis]\n",
    "    return dst\n",
    "\n",
    "X_train_normalized = (X_train - 128)/128\n",
    "\n",
    "test_img = X_train_normalized[1000]\n",
    "test_dst = random_translate(test_img)\n",
    "fig, axs = plt.subplots(1,2, figsize=(10, 3))\n",
    "\n",
    "axs[0].axis('off')\n",
    "axs[0].imshow(test_img.squeeze(), cmap='gray')\n",
    "axs[0].set_title('original')\n",
    "\n",
    "axs[1].axis('off')\n",
    "axs[1].imshow(test_dst.squeeze(), cmap='gray')\n",
    "axs[1].set_title('translated')\n",
    "\n",
    "print('shape in/out:', test_img.shape, test_dst.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Visualize the German Traffic Signs Dataset using the pickled file(s). This is open ended, suggestions include: plotting traffic sign images, plotting the count of each sign, etc. \n",
    "\n",
    "The [Matplotlib](http://matplotlib.org/) [examples](http://matplotlib.org/examples/index.html) and [gallery](http://matplotlib.org/gallery.html) pages are a great resource for doing visualizations in Python.\n",
    "\n",
    "**NOTE:** It's recommended you start with something simple first. If you wish to do more, come back to it after you've completed the rest of the sections. It can be interesting to look at the distribution of classes in the training, validation and test set. Is the distribution the same? Are there more examples of some classes than others?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "shape in/out: (32, 32, 3) (32, 32, 3)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75e6e49cf8>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def random_brightness(img):\n",
    "    shifted = img + 1.0   # shift to (0,2) range\n",
    "    img_max_value = max(shifted.flatten())\n",
    "    max_coef = 2.0/img_max_value\n",
    "    min_coef = max_coef - 0.1\n",
    "    coef = np.random.uniform(min_coef, max_coef)\n",
    "    dst = shifted * coef - 1.0\n",
    "    return dst\n",
    "\n",
    "test_dst = random_brightness(test_img)\n",
    "fig, axs = plt.subplots(1,2, figsize=(10, 3))\n",
    "\n",
    "axs[0].axis('off')\n",
    "axs[0].imshow(test_img.squeeze(), cmap='gray')\n",
    "axs[0].set_title('original')\n",
    "\n",
    "axs[1].axis('off')\n",
    "axs[1].imshow(test_dst.squeeze(), cmap='gray')\n",
    "axs[1].set_title('brightness adjusted')\n",
    "\n",
    "print('shape in/out:', test_img.shape, test_dst.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "shape in/out: (32, 32, 3) (32, 32, 1, 3)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75e6df4400>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def random_scaling(img):   \n",
    "    rows,cols,_ = img.shape\n",
    "    # transform limits\n",
    "    px = np.random.randint(-2,2)\n",
    "    # ending locations\n",
    "    pts1 = np.float32([[px,px],[rows-px,px],[px,cols-px],[rows-px,cols-px]])\n",
    "    # starting locations (4 corners)\n",
    "    pts2 = np.float32([[0,0],[rows,0],[0,cols],[rows,cols]])\n",
    "    M = cv2.getPerspectiveTransform(pts1,pts2)\n",
    "    dst = cv2.warpPerspective(img,M,(rows,cols))\n",
    "    dst = dst[:,:,np.newaxis]\n",
    "    return dst\n",
    "\n",
    "test_dst = random_scaling(test_img)\n",
    "fig, axs = plt.subplots(1,2, figsize=(10, 3))\n",
    "\n",
    "axs[0].axis('off')\n",
    "axs[0].imshow(test_img.squeeze(), cmap='gray')\n",
    "axs[0].set_title('original')\n",
    "\n",
    "axs[1].axis('off')\n",
    "axs[1].imshow(test_dst.squeeze(), cmap='gray')\n",
    "axs[1].set_title('scaled')\n",
    "\n",
    "print('shape in/out:', test_img.shape, test_dst.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "shape in/out: (32, 32, 3) (32, 32, 1, 3)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f760317f550>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def random_warp(img):\n",
    "    \n",
    "    rows,cols,_ = img.shape\n",
    "\n",
    "    # random scaling coefficients\n",
    "    rndx = np.random.rand(3) - 0.5\n",
    "    rndx *= cols * 0.06   # this coefficient determines the degree of warping\n",
    "    rndy = np.random.rand(3) - 0.5\n",
    "    rndy *= rows * 0.06\n",
    "\n",
    "    # 3 starting points for transform, 1/4 way from edges\n",
    "    x1 = cols/4\n",
    "    x2 = 3*cols/4\n",
    "    y1 = rows/4\n",
    "    y2 = 3*rows/4\n",
    "\n",
    "    pts1 = np.float32([[y1,x1],\n",
    "                       [y2,x1],\n",
    "                       [y1,x2]])\n",
    "    pts2 = np.float32([[y1+rndy[0],x1+rndx[0]],\n",
    "                       [y2+rndy[1],x1+rndx[1]],\n",
    "                       [y1+rndy[2],x2+rndx[2]]])\n",
    "\n",
    "    M = cv2.getAffineTransform(pts1,pts2)\n",
    "\n",
    "    dst = cv2.warpAffine(img,M,(cols,rows))\n",
    "    \n",
    "    dst = dst[:,:,np.newaxis]\n",
    "    \n",
    "    return dst\n",
    "\n",
    "test_dst = random_warp(test_img)\n",
    "\n",
    "fig, axs = plt.subplots(1,2, figsize=(10, 3))\n",
    "\n",
    "axs[0].axis('off')\n",
    "axs[0].imshow(test_img.squeeze(), cmap='gray')\n",
    "axs[0].set_title('original')\n",
    "\n",
    "axs[1].axis('off')\n",
    "axs[1].imshow(test_dst.squeeze(), cmap='gray')\n",
    "axs[1].set_title('warped')\n",
    "\n",
    "print('shape in/out:', test_img.shape, test_dst.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X, y shapes: (34799, 32, 32, 3) (34799,)\n",
      "0 : \n",
      "1 : \n",
      "2 : \n",
      "3 : \n",
      "4 : \n",
      "5 : \n",
      "6 : \n",
      "7 : \n",
      "8 : \n",
      "9 : \n",
      "10 : \n",
      "11 : \n",
      "12 : \n",
      "13 : \n",
      "14 : \n",
      "15 : \n",
      "16 : \n",
      "17 : \n",
      "18 : \n",
      "19 : \n",
      "20 : \n",
      "21 : \n",
      "22 : \n",
      "23 : \n",
      "24 : \n",
      "25 : \n",
      "26 : \n",
      "27 : \n",
      "28 : \n",
      "29 : \n",
      "30 : \n",
      "31 : \n",
      "32 : \n",
      "33 : \n",
      "34 : \n",
      "35 : \n",
      "36 : \n",
      "37 : \n",
      "38 : \n",
      "39 : \n",
      "40 : \n",
      "41 : \n",
      "42 : \n",
      "X, y shapes: (34799, 32, 32, 3) (34799,)\n"
     ]
    }
   ],
   "source": [
    "X_train_normalized = (X_train - 128)/128\n",
    "print('X, y shapes:', X_train_normalized.shape, y_train.shape)\n",
    "\n",
    "input_indices = []\n",
    "output_indices = []\n",
    "\n",
    "for class_n in range(n_classes):\n",
    "    print(class_n, ': ', end='')\n",
    "    class_indices = np.where(y_train == class_n)\n",
    "    n_samples = len(class_indices[0])\n",
    "    if n_samples < 50:\n",
    "        for i in range(50 - n_samples):\n",
    "            input_indices.append(class_indices[0][i%n_samples])\n",
    "            output_indices.append(X_train_normalized.shape[0])\n",
    "            new_img = X_train_normalized[class_indices[0][i % n_samples]]\n",
    "            new_img = random_translate(random_scaling(random_warp(random_brightness(new_img))))\n",
    "            X_train_normalized = np.concatenate((X_train_normalized, [new_img]), axis=0)\n",
    "            y_train = np.concatenate((y_train, [class_n]), axis=0)\n",
    "            #if i % 50 == 0:\n",
    "            #    print('|', end='')\n",
    "            #elif i % 10 == 0:\n",
    "            #    print('-',end='')\n",
    "    print('')\n",
    "            \n",
    "print('X, y shapes:', X_train_normalized.shape, y_train.shape)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n"
     ]
    },
    {
     "ename": "ValueError",
     "evalue": "low >= high",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-83-a8b052e304ea>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0mpicks\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mi\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m5\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 5\u001b[0;31m     \u001b[0mrnd_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrandom\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrandint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlow\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mhigh\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mchoices\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      6\u001b[0m     \u001b[0mpicks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mchoices\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpop\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrnd_index\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      7\u001b[0m \u001b[0mfig\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msubplots\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;36m5\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfigsize\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m15\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m6\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32mmtrand.pyx\u001b[0m in \u001b[0;36mmtrand.RandomState.randint (numpy/random/mtrand/mtrand.c:16117)\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;31mValueError\u001b[0m: low >= high"
     ]
    }
   ],
   "source": [
    "choices = list(range(len(input_indices)))\n",
    "print(len(input_indices))\n",
    "picks = []\n",
    "for i in range(5):\n",
    "    rnd_index = np.random.randint(low=0,high=len(choices))\n",
    "    picks.append(choices.pop(rnd_index))\n",
    "fig, axs = plt.subplots(2,5, figsize=(15, 6))\n",
    "fig.subplots_adjust(hspace = .2, wspace=.001)\n",
    "axs = axs.ravel()\n",
    "for i in range(5):\n",
    "    image = X_train_normalized[input_indices[picks[i]]].squeeze()\n",
    "    axs[i].axis('off')\n",
    "    axs[i].imshow(image, cmap = 'gray')\n",
    "    axs[i].set_title(y_train[input_indices[picks[i]]])\n",
    "for i in range(5):\n",
    "    image = X_train_normalized[output_indices[picks[i]]].squeeze()\n",
    "    axs[i+5].axis('off')\n",
    "    axs[i+5].imshow(image, cmap = 'gray')\n",
    "    axs[i+5].set_title(y_train[output_indices[picks[i]]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f76030be898>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# histogram of label frequency\n",
    "hist, bins = np.histogram(y_train, bins=n_classes)\n",
    "width = 0.8 * (bins[1] - bins[0])\n",
    "center = (bins[:-1] + bins[1:]) / 2\n",
    "plt.bar(center, hist, align='center', width=width)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[180, 1980, 2010, 1260, 1770, 1650, 360, 1290, 1260, 1320, 1800, 1170, 1890, 1920, 690, 540, 360, 990, 1080, 180, 300, 270, 330, 450, 240, 1350, 540, 210, 480, 240, 390, 690, 210, 599, 360, 1080, 330, 180, 1860, 270, 300, 210, 210]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f760307a828>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "### Data exploration visualization code goes here.\n",
    "### Feel free to use as many code cells as needed.\n",
    "import matplotlib.pyplot as plt\n",
    "import random\n",
    "# Visualizations will be shown in the notebook.\n",
    "%matplotlib inline\n",
    "\n",
    "# Visualization\n",
    "#training data \n",
    "hist_train_y = []\n",
    "hist_train_x = []\n",
    "\n",
    "train_y_list = y_train.tolist()\n",
    "\n",
    "for i in range(0, n_classes):\n",
    "    hist_train_x.append(i)\n",
    "    count = train_y_list.count(i)\n",
    "    hist_train_y.append(count)\n",
    "b_width = 0.9\n",
    "fig, ax = plt.subplots()\n",
    "rects1 = ax.bar(hist_train_x, hist_train_y, b_width, label='Labels')\n",
    "ax.set_xticks(hist_train_x)\n",
    "ax.set_xticklabels(train_y_list)\n",
    "\n",
    "#plt.plot.bar(hist_train_x, hist_train_y, 'b-')\n",
    "print (hist_train_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7602f734e0>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.xlabel('Labels')\n",
    "plt.ylabel('Frequency')\n",
    "plt.title('Training Data Set')\n",
    "plt.axis([0, 43, 0, 2200])\n",
    "plt.grid(True)\n",
    "\n",
    "plt.plot(hist_train_x, hist_train_y, 'b-')\n",
    "plt.show()\n",
    "#print (hist_train_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X Train Normalized - mean 1.21925099939\n",
      "X Train Normalized - mean 1.21879869422\n"
     ]
    }
   ],
   "source": [
    "## Normalize the train and test datasets to (-1,1)\n",
    "\n",
    "X_train_normalized = (X_train - 128)/128 \n",
    "X_test_normalized = (X_test - 128)/128\n",
    "\n",
    "print(\"X Train Normalized - mean\", np.mean(X_train_normalized))\n",
    "print(\"X Train Normalized - mean\", np.mean(X_test_normalized))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "----\n",
    "\n",
    "## Step 2: Design and Test a Model Architecture\n",
    "\n",
    "Design and implement a deep learning model that learns to recognize traffic signs. Train and test your model on the [German Traffic Sign Dataset](http://benchmark.ini.rub.de/?section=gtsrb&subsection=dataset).\n",
    "\n",
    "The LeNet-5 implementation shown in the [classroom](https://classroom.udacity.com/nanodegrees/nd013/parts/fbf77062-5703-404e-b60c-95b78b2f3f9e/modules/6df7ae49-c61c-4bb2-a23e-6527e69209ec/lessons/601ae704-1035-4287-8b11-e2c2716217ad/concepts/d4aca031-508f-4e0b-b493-e7b706120f81) at the end of the CNN lesson is a solid starting point. You'll have to change the number of classes and possibly the preprocessing, but aside from that it's plug and play! \n",
    "\n",
    "With the LeNet-5 solution from the lecture, you should expect a validation set accuracy of about 0.89. To meet specifications, the validation set accuracy will need to be at least 0.93. It is possible to get an even higher accuracy, but 0.93 is the minimum for a successful project submission. \n",
    "\n",
    "There are various aspects to consider when thinking about this problem:\n",
    "\n",
    "- Neural network architecture (is the network over or underfitting?)\n",
    "- Play around preprocessing techniques (normalization, rgb to grayscale, etc)\n",
    "- Number of examples per label (some have more than others).\n",
    "- Generate fake data.\n",
    "\n",
    "Here is an example of a [published baseline model on this problem](http://yann.lecun.com/exdb/publis/pdf/sermanet-ijcnn-11.pdf). It's not required to be familiar with the approach used in the paper but, it's good practice to try to read papers like these."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Pre-process the Data Set (normalization, grayscale, etc.)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Minimally, the image data should be normalized so that the data has mean zero and equal variance. For image data, `(pixel - 128)/ 128` is a quick way to approximately normalize the data and can be used in this project. \n",
    "\n",
    "Other pre-processing steps are optional. You can try different techniques to see if it improves performance. \n",
    "\n",
    "Use the code cell (or multiple code cells, if necessary) to implement the first step of your project."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X Train RGB shape: (34799, 32, 32, 3)\n",
      "X Test RGB shape: (12630, 32, 32, 3)\n",
      "X Validate RGB shape: (4410, 32, 32, 3)\n",
      "X Train Grayscale shape: (34799, 32, 32)\n",
      "X Test Grayscale shape: (12630, 32, 32)\n",
      "X Valid Grayscale shape: (4410, 32, 32)\n",
      "1.21925099939\n",
      "1.21879869422\n",
      "Original shape: (34799, 32, 32, 3)\n",
      "Normalized shape: (34799, 32, 32, 3)\n",
      "18\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f754c608400>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Import the required processing modules\n",
    "\n",
    "import random\n",
    "\n",
    "#image_x_shape = X_train.shape\n",
    "\n",
    "### Preprocess the data here. It is required to normalize the data. Other preprocessing steps could include \n",
    "### converting to grayscale, etc.\n",
    "### Feel free to use as many code cells as needed.\n",
    "def conv_rgb2gray(rgb):\n",
    "    return np.dot(rgb, [0.299, 0.587, 0.114])\n",
    "\n",
    "print('X Train RGB shape:', X_train.shape)\n",
    "print('X Test RGB shape:', X_test.shape)\n",
    "print('X Validate RGB shape:', X_valid.shape)\n",
    "\n",
    "\n",
    "if (X_train.shape[3] == 3):\n",
    "    # Grayscale conversion of image\n",
    "    X_train_gray = conv_rgb2gray(X_train)\n",
    "    X_valid_gray = conv_rgb2gray(X_valid)\n",
    "    X_test_gray = conv_rgb2gray(X_test)\n",
    "    print('X Train Grayscale shape:', X_train_gray.shape)\n",
    "    print('X Test Grayscale shape:', X_test_gray.shape)\n",
    "    print('X Valid Grayscale shape:', X_valid_gray.shape)\n",
    "    \n",
    "    \n",
    "    from sklearn import preprocessing\n",
    "    \n",
    "    for i, picture in enumerate(X_train_gray):\n",
    "        X_train_gray[i] = preprocessing.normalize(picture, norm='l2', axis=1, copy=True, return_norm=False)\n",
    "    for i, picture in enumerate(X_valid_gray):\n",
    "        X_valid_gray[i] = preprocessing.normalize(picture, norm='l2', axis=1, copy=True, return_norm=False)\n",
    "    for i, picture in enumerate(X_test_gray):\n",
    "        X_test_gray[i] = preprocessing.normalize(picture, norm='l2', axis=1, copy=True, return_norm=False)\n",
    "\n",
    "    # Another way to normalize\n",
    "    X_train_normalized = (X_train - 128)/128 \n",
    "    X_test_normalized = (X_test - 128)/128\n",
    "    print(np.mean(X_train_normalized))\n",
    "    print(np.mean(X_test_normalized))\n",
    "    \n",
    "    print(\"Original shape:\", X_train.shape)\n",
    "    print(\"Normalized shape:\", X_train_normalized.shape)\n",
    "    \n",
    "    \n",
    "    # Reshape Grayscale Pictures (Add Dimension 1)\n",
    "    X_train_norm = X_train_gray.reshape(X_train.shape[0], X_train.shape[1], X_train.shape[2], 1)\n",
    "    X_valid_norm = X_valid_gray.reshape(X_valid.shape[0], X_valid.shape[1], X_valid.shape[2], 1)\n",
    "    X_test_norm = X_test_gray.reshape(X_test.shape[0], X_test.shape[1], X_test.shape[2], 1)\n",
    "\n",
    "    # Print one random sample image from the training set and the corresponding label\n",
    "    index = random.randint(0, len(X_train)-1)\n",
    "    image = X_train[index].squeeze()\n",
    "\n",
    "    # Print \"before picture\"\n",
    "    plt.figure(figsize=(2,2))\n",
    "    plt.imshow(image, cmap=\"gray\")\n",
    "    print(y_train[index])\n",
    "    \n",
    "    X_train = X_train_norm\n",
    "    X_valid = X_valid_norm\n",
    "    X_test = X_test_norm\n",
    "    \n",
    "    # Save Image Depth in Variable\n",
    "    image_depth = X_train.shape[3]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RGB shape: (34799, 32, 32, 1)\n",
      "Grayscale shape: (34799, 32, 32, 1)\n"
     ]
    }
   ],
   "source": [
    "### Preprocess the data here.\n",
    "### Feel free to use as many code cells as needed.\n",
    "\n",
    "# Convert to grayscale\n",
    "X_train_rgb = X_train\n",
    "X_train_gry = np.sum(X_train/3, axis=3, keepdims=True)\n",
    "\n",
    "X_test_rgb = X_test\n",
    "X_test_gry = np.sum(X_test/3, axis=3, keepdims=True)\n",
    "\n",
    "print('RGB shape:', X_train_rgb.shape)\n",
    "print('Grayscale shape:', X_train_gry.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Completed Preprocessing\n"
     ]
    }
   ],
   "source": [
    "X_train = X_train_gry\n",
    "X_test = X_test_gry\n",
    "\n",
    "print('Completed Preprocessing')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7602e7b5f8>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Print one random sample image from the training set and the corresponding label\n",
    "#index = random.randint(0, len(X_train)-1)\n",
    "\n",
    "# Visualize rgb vs grayscale\n",
    "n_rows = 8\n",
    "n_cols = 10\n",
    "offset = 9000\n",
    "fig, axs = plt.subplots(n_rows,n_cols, figsize=(18, 14))\n",
    "fig.subplots_adjust(hspace = .1, wspace=.001)\n",
    "axs = axs.ravel()\n",
    "for j in range(0,n_rows,2):\n",
    "    for i in range(n_cols):\n",
    "        index = i + j*n_cols\n",
    "        image = X_train_rgb[index + offset]\n",
    "        axs[index].axis('off')\n",
    "        #axs[index].imshow(image)\n",
    "    for i in range(n_cols):\n",
    "        index = i + j*n_cols + n_cols \n",
    "        image = X_train_gry[index + offset - n_cols].squeeze()\n",
    "        axs[index].axis('off')\n",
    "        axs[index].imshow(image, cmap='gray')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Un-Shuffled \n",
      "\n",
      "[41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41\n",
      " 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41\n",
      " 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41\n",
      " 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41\n",
      " 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41\n",
      " 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41\n",
      " 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41\n",
      " 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41\n",
      " 41 41 41 41 41 41 41 41 41 41 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31\n",
      " 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31\n",
      " 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31\n",
      " 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31\n",
      " 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31\n",
      " 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31\n",
      " 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31\n",
      " 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31\n",
      " 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31\n",
      " 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31\n",
      " 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31\n",
      " 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31 31]\n",
      "\n",
      "Shuffling completed \n",
      "\n",
      "[ 4 16 17 17  2 38  2 42  5 23  8  9  8 23 31  8  2 31  1 28  8 13  4 29 17\n",
      "  5 17 22 12 10 10 12 12  9  1  1 18  9 15 12 12 17  2 13  5 13 35  1 35 30\n",
      " 12 14 11  3 35 17  9 26 36  5 17 17 35  2 35 33 13 26  4 38 41 27 23 16  2\n",
      " 10 31 15  5 25 25 41  8  9 35  3 35  2  3 10 15 13  2 14  6  7 12  5 12 10\n",
      "  4 13 10 12  2 11  8 38 17 18  3 38  2  2 42  7 25 12 10  1  2 11  2 12  9\n",
      " 30  3  6 13 27 13  3  7 42 26 25  2  5  0 25 10 21  2 10 33 38 15  1 12  7\n",
      " 35 27 18  9 10  5  1 10  7 12  4 20  5 13  1  1  4 18  1  3  8  9  3 12 13\n",
      " 18 39 42 35 38 31  5 25 25  8  5  8  2  5  2 23 13 10 12  9 20 35  3 21  9\n",
      "  2  9  2 27  2 25 16  2 20  5  9 24  1  2 25  3 18  1  2 24 15 15 10  9 15\n",
      "  1 41 30 38  1 17 33 33  5 35  1  1  4 35  1 41 15  4 11  3 18  7  3  3  2\n",
      " 38  1 12 25  2  4 16 17 11  5  7 22 38 30 10 26  1  3  2 13 31 18 17  8 40\n",
      " 16  7  8 38  8 31 13  5 12  5  9 12 14 11 13  2 11 13  5 23 20 33 39 11 40\n",
      "  4  4  9  8  3  8  5 12 13 23 17 36 32 38 11  2  4  7 19 34  7 30 10  3  9\n",
      "  1  1 16 28 12  5 22 13 33 25 25 25 12 14  1  8 16  8 21  1 13  4 33 22 12\n",
      "  5  4 35  4  2  5 18 11  4  4 23  8 25  7  3  7 42  1 18 24 17 35 10  2  8\n",
      " 31 11 18 10  1  1 33  5 25 40  8 15  7  3 32  1  5 14 25 15 16 28 10 12 38\n",
      " 35 11  4  1  9  4 38  4 26 10 33  1 25 38  8  9  4 23  4 37 38 30  1 14 38\n",
      " 10 22 38  5  5  4 14 11 18  9 14 37  9 33  9 38 38 12 22 17  4 25 38 30  7\n",
      " 40 38  1 12 12 17  3 13 18 13 18 38 16  5 32  4  3  9 11 40 28  3 30 40  9\n",
      "  6 35 25 25  4  7 16 38  7  1 13 41  1  2  7 28  3  1  2 11 15 28 12  3  8]\n"
     ]
    }
   ],
   "source": [
    "### Preprocess the data here. Preprocessing steps could include normalization, converting to grayscale, etc.\n",
    "### Feel free to use as many code cells as needed.\n",
    "\n",
    "print(\"\\nUn-Shuffled \\n\")\n",
    "print(y_train[0:500])\n",
    "\n",
    "# shuffle the data\n",
    "from sklearn.utils import shuffle\n",
    "\n",
    "X_train, y_train = shuffle(X_train, y_train)\n",
    "\n",
    "print(\"\\nShuffling completed \\n\")\n",
    "print(y_train[0:500])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.0546461022947\n",
      "0.0542052554114\n"
     ]
    }
   ],
   "source": [
    "print(np.mean(X_train))\n",
    "print(np.mean(X_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-0.999573077326\n",
      "-0.999576521442\n"
     ]
    }
   ],
   "source": [
    "## Normalize the train and test datasets to (-1,1)\n",
    "\n",
    "X_train_normalized = (X_train - 128)/128 \n",
    "X_test_normalized = (X_test - 128)/128\n",
    "\n",
    "print(np.mean(X_train_normalized))\n",
    "print(np.mean(X_test_normalized))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [],
   "source": [
    "def random_translate(img):\n",
    "    rows,cols,_ = img.shape\n",
    "    \n",
    "    # allow translation up to px pixels in x and y directions\n",
    "    px = 2\n",
    "    dx,dy = np.random.randint(-px,px,2)\n",
    "\n",
    "    M = np.float32([[1,0,dx],[0,1,dy]])\n",
    "    dst = cv2.warpAffine(img,M,(cols,rows))\n",
    "    \n",
    "    dst = dst[:,:,np.newaxis]\n",
    "    \n",
    "    return dst"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Original shape: (34799, 32, 32, 1)\n",
      "Normalized shape: (34799, 32, 32, 1)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f75fb048898>"
      ]
     },
     "execution_count": 96,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75fb1d9e80>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfkAAADSCAYAAABJqrmlAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4wLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvpW3flQAAGCJJREFUeJzt3XuQ1eV9x/HPF1hYcHeB5X6vAolKQKpBKUao1EYnl9qMbZNM2ow2aZM2naaZJJ10YpJJm3Zyc9KZTnpNYtrEGjudidGJzSSaqTNWFBANQauJglwWWNjltrDLLsrTP34/skee7w+e5cBh9+n7NePM8t1zfrdzfL772/PZ57EQggAAQH5GXewDAAAAFwZNHgCATNHkAQDIFE0eAIBM0eQBAMgUTR4AgEzR5DNhZi+Z2a+c78eeZTsfNLOH690OAJwrM2s2s2Bmc8t/f9PM/vw872PEjnU0+TMws6M1/500s76af7+nju0+YWa/ez6PNYSwMISw7nw/FsDwMZLGpIslhHB7COGLF/s4hosxF/sAhrMQQsupr83sZUnvDyEMq5/mzGxMCOGVi30cAC68kTAmnQ1jVmNxJ18HMxttZp8ys61m1mVm95jZpPJ7l5jZd8zsgJkdMrMnzWyymd0laYWkr5U/fd9Vse3bzOy58rkPm9nimu/tNbOPmdmzko7U1N5Uft1iZv9ePneLmf2Fmb142vNPPfbz5XHfa2Y9ZrbZzJbXPPbTZrat/N4WM3vrBbiUAM6DCzUmmdnlZvaKmd1hZrvMbL+Zfbzm++PN7Ktmtqf8/pfMrKn83i1m9mJ5XJ2S/qGmdmd5nB1m9hYzu7X8OLHbzD5as/3ry+M9bGa7zewrZubepJbneGf59Q+d3368q/zeG8zsx2Z20Mz+18x+s2Yb083sITM7YmbrJC04H6/PxUCTr8/HJb1Z0pskzZV0QtJXyu+9X8VvSuZImirpTyQNhBA+KmmDip/AW8p/v4aZvUHSNyX9saTpkh6V9MBpb+p3Svp1SVOc4/qcpGkq3phvlfR7ZzmPd0j6hqRJkh6R9Lc133tB0ipJEyV9QdJ3zGzqWbYH4OK4IGNSabSkN0paJOktkv7azC4rv/dZScskLZV0jaRflVT7ufgvSWqSNE/Sn5a1BeXxzZT0eRVj0G+V27mp3P6c8rEnyuNtl3SDpLeX53NGIYQ3l+fUIuk9knZLetTM2iT9SNLXy2vxXknfMLNF5VP/WdIBSTMk/ZGk3z/bvoYrmnx9PiDpEyGE3SGE4yre6O80M1PxppwmaWEI4ZUQwoYQwrHE7b5b0ndDCP8dQhiQ9Dcq3ohvrHnMV8r99jnP/x1JnwshHA4hbJf092fZ349DCD8KIbwq6VuSfnEnH0K4L4SwJ4RwMoTwLUkdKv4nBjD8XKgx6ZTPhBCOhxA2SHpeRUOWigb6mRBCVwihU8WNRu3NRb+kvwohDNSMWb2SvlT+6v47Khrql0MIx0IIT0t6ScUPDQohrC+P99UQwkuSviZpTepBm9mS8jm3hRD2qLix2RJCuKfc5gZJD0q6zcyaJf2GpDtDCH0hhGck3TPE6zRs8Jn8OSr/p5kn6SEzq13lZ5SKu+uvq/gJ9T/NrEXSv0n6VNlIz2a2pO2n/hFCeNXMOlT8BH7KzjMc14zTvu8+tsbemq97JdV+7vc+SR+WNL8staj4gQPAMHKBxyRJejWE0FXz715JLeV+Z6pmzCq/rh2v9oYQTpy2vf0hhJPl16caf2fN9/tUjkVmdqWkuyRdLWm8it71PykHbWbtkr4n6WMhhCfL8gJJq83sUM1Dx0g6WJ6L6bXj5nYN/kAzonAnf45CsXxfh6S1IYRJNf81lz/N9ocQPh1CuFzSakm/Leldp55+ls3vVs1nQGY2WsX/MB21h3CG49qn4ld1p8wbyrnV7Pd1kv5O0h9Kag8hTJL0oor/AQAMIxd4TDrbfvfqtZ9bz1fCeDUE/yJpk4rfQrRJ+ksljEPl2HmfpAdCCP9a862dkn542nVqCSH8WXkuQa8dN+drhKLJ1+cfJX3ezOZJvwhrvL38+iYzu9LMRqkIx70i6dRPzJ2SLvM2WLpP0jvMbHUZXvmEpG5JGxOP6z8kfdLMJprZfBWfKZ2LFkknJe2XNMrMPqji8zgAw9OFGpPO5l5JnzGzKWY2XdInJX27ju2drlXS4RDC0fJX73+Q+Lwvq2jYHz+tfr+kXzazd5pZk5mNNbOVZva68mOOByV9tgwULlPxccSIRJOvzxclPSzpx2bWI+lxFb9Okoo77+9J6pG0RdJDKpqvVARh3lumOqO/5wwhbJb0Pkn/pKLB/pqkW4fwZyd3qvi103ZJ/1Xut3+oJxdC2KRi0NgoaY+kS5X+gwaAxrsgY1KCT0t6TtKzkp5R8av08/m36h+R9H4zOyrpqypuhFK8W0VQ73BNwv62EMJBSTdLukPF2LZbRY6gqXzeB1R87NmpYhy++7ydSYNZ8ZsW5MzMPiLplhDCzRf7WAAAjcOdfIbMbF75q6dR5a+2Pizpuxf7uAAAjUW6Pk/jVPzN6QIVv7b/too/HwEA/D/Cr+sBAMgUv64HACBTNHkAADLV0M/k77rrruizgZMnT0aPKyZQOnevvpo2gZP3UYV3PF6ttbU16XFVx+LVR42Kf+byan198Uy2o0ePdveTsj3vuCWpvz/+q7vU6+O9ht45v/JK/FeB+/fvj2qTJ092j9HbZlNTU1Tr7u6OanfffTeT+iAZ49eZ64xfheE2fnEnDwBApmjyAABkiiYPAECmGvqZfG9vb1TzPldK/Tyk6s//xoyJT+vEidMXQPKlHk9PT0/Sc+v9E8XUz6C86+Md91A+Lxw7dmzS81M/Q0z9zG7q1HiRu4GBgaR9VGlra6vr+QDj19AxfhUu5vjFnTwAAJmiyQMAkCmaPAAAmaLJAwCQqYYG7+oJqXgBh6oQhjdBgfd8Tz0TTKSeX1V9KJNRnOv2vMdVTURRzzl64aHUY/T2kfr6Ve0HqBfj15nrjF/V+7iY4xd38gAAZIomDwBApmjyAABkiiYPAECmGhq888IMqbyAQ1WYwQtipAYfUkMvqTVve5IfSPFmRfJmukoNs3irGXm1qqBH6utVzwpS3r6HElLyDCVABKRi/BrE+FW97+E2fnEnDwBApmjyAABkiiYPAECmaPIAAGSqocG71tbWqOYFKTxeAGQoSzVWBUhSeIESbxnDvr6+qHbkyBF3mzt27Ihq27Zti2pdXV1Rrb+/393m6bxrO3/+/Kh2xRVXuM+fPXt2VJs4cWJUa25ujmqps36lhlS84FJV3XtfpL7PgCqMX4MYv6prw2384k4eAIBM0eQBAMgUTR4AgEzR5AEAyBRNHgCATDU0XT99+vT4AJwkqZcG9RKLVYnT1GkKPd6+jx8/HtV27doV1Z566qmo9uyzz7r7OXr0aFTzUpWpUxx659zb25t0PFXH6KWJr7rqqqh20003RbU5c+ZENS+xmvpaedNjSn7q1NsP6XrUi/FrEONXYSSMX9zJAwCQKZo8AACZoskDAJApmjwAAJlqaPCura0tqnkhDC8w4QU4qsIM3pSN3vSB3vN3794d1Z5++umo9sQTT0S1zs7OqFYVrvGmQ/SO0bs+Xq1qisyUx1U999ChQ1Ht0UcfjWoHDx6MajfccENUW7lyZVTzXmsveFJ1jN519B6buh43UIXxaxDjV2EkjF+MfAAAZIomDwBApmjyAABkiiYPAECmGhq889QTwqiaBch7vhcK6ejoiGrr1q2Lao899lhU89ZZ9s7FC9FI/uxZl19+eVSbNWtWVGtvb49qXthj586dUW3r1q1RzVsHWvJntfL2s3nz5qjmhV689arXrl0b1caNGxfVqt4TqWEf4EJg/BrE+FUYbuMXd/IAAGSKJg8AQKZo8gAAZIomDwBAphoavOvq6opqqSGD1DDLUPbtzfq0fv36qHbgwIGkfcyfPz+qXX311e5jlyxZEtVmzJgR1Zqbm5P27bnyyiujmhf02L59u/t8L7DjXTNvViwvDPODH/wgqnmzdl177bXu8XhSZ9SqZ/lOQGL8qsX4VRgJ4xd38gAAZIomDwBApmjyAABkiiYPAECmGhq88wIO3gxEXvDAC1xU8WYm+ulPfxrVNm3aFNW8kIoXerjiiiui2nXXXRfVFi9e7B5jaiDl+PHjUa2eGZG8WbbmzZvnPvbWW2+NanPnzo1qDz30UFTzZozyZrDywjFegMebYUvyz8d7r7DULOrF+DWI8aswEsYvRj4AADJFkwcAIFM0eQAAMkWTBwAgUw0N3nkhldTHDSW4snv37qjmBVc6OzujmhdSWbBgQVTzZjW67LLLotqYMf4lruccvWP0ghne9oayX2/JxOXLl0c1L5D0wAMPRDUvhOPNVuXN2nXjjTe6x9ja2hrVvNnFvGMEhoLxaxDjV2EkjF/cyQMAkCmaPAAAmaLJAwCQKZo8AACZamjwzgtXDAwMJD3OCyN4y/xJ0o4dO6JaR0dH0janTp0a1a655pqotnDhwqhW73Km3vl420wNn3jP9R5XNQOV91gvzOItCblr166otmHDhqjmvf5btmyJat4MXZI0fvz4qFbvjGOAh/HrzBi/CsNt/OJOHgCATNHkAQDIFE0eAIBM0eQBAMhUQ4N37e3tUc0Lj6QGM/bt2+fux6t7MwZNmDAhqnkhjBUrVkS1adOmRbXUwI0k9fT0RLWtW7dGNW/pyKNHj0Y1L8DhnZ8XPBk7dqx7jN5jvW16+161alVU88JDXsDl2LFjUc07Z0lqa2uLat5xA/Vi/BrE+FUYCeMXd/IAAGSKJg8AQKZo8gAAZIomDwBApmjyAABkqqHp+qVLl0a11KkivcRiVfLTm17RS1V6CVMvVelNC3n48OGo5q0D7a1BLEldXV1Rbdu2bVHtvvvui2reNaua2jHluVXX0XtsU1NTVFuyZElU+9CHPhTVvLWcvevonUvVMS5atCiqeVN7AvVi/BrE+FUYCeMXd/IAAGSKJg8AQKZo8gAAZIomDwBAphoavPOm6/MCCV44wltb2AuzSFJvb29U88IQs2fPjmrz5s2Lat6Ukl64ZsyY+HJ+//vfd4+xubk5qu3Zs8d97Om8aTO9kIl3PN719p4r+efoPd8L53jn561hvX79+qTtHTp0yD1GjzfNpfeeAoaC8WsQ41dhJIxf3MkDAJApmjwAAJmiyQMAkCmaPAAAmWpo8M4LUnjrLHs1L6xx5MgRdz9e8MHb9yWXXBLVvJmlvNmd+vr6opoXmPACHFJ6GCY1aOIFc1JnXvKut5QefPHOsaWlJap5Mzl518x7/apCSt41894rVWtOA6kYvwYxfhVGwvjFnTwAAJmiyQMAkCmaPAAAmaLJAwCQqYYG76ZMmRLVvACHFzzwZvypCiN4dW/5xtbW1qjmhTWeeeaZqHbvvfdGNW9Wo6pjvO6666KaFwAZP358VOvv749qqdfHC6lULYPo7cc7xokTJ0a1yZMnR7WDBw9GNe/8Umegqtr3zJkzo5r3ugJDwfg1iPGrMBLGL+7kAQDIFE0eAIBM0eQBAMgUTR4AgEw1NI20c+fOqJYaMvFmDPKWZJT8WYS8EIYXNOns7Ixqhw8fjmpe4Oaqq66Kal4IQ5K2b98e1WbNmhXVbr/99qiWOuuTFwDyalXX0ZuRy7uO3pKXR48ejWo7duyIat7r6r1+Vbq7u6Pavn37opp3febOnZu8H4DxaxDjV2EkjF/cyQMAkCmaPAAAmaLJAwCQKZo8AACZamjwzlsa0VtO0Jv9yOMt/Sf5ywR6Syt6wZVx48ZFtSVLlkS1jRs3RrXNmzdHtWXLlrnH6M2etX///qjmhT28c/HCHu3t7UmP84JCkj/rU+oykd5r6F1vLwA0evToqDZ9+nT3GL3Zodra2qJa1TkCqRi/BjF+FUbC+MWdPAAAmaLJAwCQKZo8AACZoskDAJCphgbvHn/88ahWtUzg6bzggTczkOTPgOSFJn72s59FtQ0bNkQ1L3yyePHiqOYFZrwlGSVp0aJFUc1bttCb6Sg1hLFt27ao9thjj0U17zpI0rx586LaihUropq3XOKTTz6ZtG8vrOPNIubNLCVJTz31VFSbMGFCVPMCN29729vcbQIexq9BjF+FkTB+cScPAECmaPIAAGSKJg8AQKZo8gAAZKqhwTsvaOIFV7wZiLwlBr3ZhiQ/uOA93wuzbNq0Kap5S/pde+21Uc0LmXizH0l+uKZqWccU3nWcNm1aVLv55puj2urVq91terM+eeEa71y8AJC3PKUXwmltbY1q3uxVknTgwIGo5oVcvPcUMBSMX4MYvwojYfziTh4AgEzR5AEAyBRNHgCATNHkAQDIFE0eAIBMNTRd703N5yU6U1Oe3vrOknTppZdGNS8Z2dPTE9VefvnlqOYlVtesWRPVvDRnVSrSuxZewtS7Fqnb82rNzc1RzVuDuur5Xsr3pZdeimrPP/98VDty5EhU866Z9/p5SVvJn0LSu+ap1xGowvg1iPGrMBLGL+7kAQDIFE0eAIBM0eQBAMgUTR4AgEw1NHg3Zky8u9TgihdG8EIPkvT6178+qv385z+Pai+++GJU88IVGzdujGreNIXLly+Pal6wQkoPrqQ+zpN6zapCHV5969atUc27Prt27Uo5RDeQ4q1/7a35LKUHdrz3HjAUjF+DGL8KI2H84k4eAIBM0eQBAMgUTR4AgEzR5AEAyFRD00jerE9eoMBbZ9kLYVStdeyFSlatWhXVurq6olp3d3dU89aRfuSRR6LawMBAVLv++uvdY/R45+iFMFJ5z/WurTcLlCQ999xzUW3dunVR7YUXXohqXujFm+HLC/ssWrQoqnkzXUn+OXrvi6pzBFIxfp0Z41dhuI1f3MkDAJApmjwAAJmiyQMAkCmaPAAAmWpo8M4LH3izH40fPz6qpc4sJflhGG8WqRtvvDGqeYGUQ4cORbW9e/dGtQcffDCqebMpSdItt9wS1ZYsWRLVWlpaopp33t519II03kxOXhhFkrZs2RLVDhw4ENW8wI23/KMX4lm5cmVUmzVrVtI+quoE73AhMH4NYvwqjITxizt5AAAyRZMHACBTNHkAADJFkwcAIFMNDd7NnDkzqlWFT1J4sx9Jfohj0qRJUc1bRrGtrS2qPfzww1Gts7MzqnnhiI6ODvcY77nnnqR9ezUvzOKFVLxlJ3t6eqLasWPH3GP0zscLH3nXds2aNVFt9erVUW3BggVRzQu9VAVPvLo3i1Q97zNAYvyqxfhVGAnjF3fyAABkiiYPAECmaPIAAGSKJg8AQKYaGrzzghQeL3gyFF6Y4cSJE1HNC4B4SwdOmTIlqt1///1RzZtFqq+vzz1GLyzS29ubtE1vlqSqGZVOlzrblCQ1NTVFNW82p7Vr10a1ZcuWRTXvOvb39yfVqqS+V1KvD1CF8WsQ41dhJIxf3MkDAJApmjwAAJmiyQMAkCmaPAAAmWpo8O7gwYNRzQseeCED73FVoQUvuOIFNrzHecs8zpkzJ6rdcccdUe0nP/lJVKtaqtELpHjhGu8YU6+P9zhvxqcZM2a4x+gtHbl06dKoNm3atKjmXUdvyUvvGL3Xytte1WO9bXqzSAFDwfg1iPGr+hiH2/jFyAcAQKZo8gAAZIomDwBApmjyAABkqqHBuwMHDkQ1b7lFL2TgBThGjx7t7scLcXj78WpeQKJqP6dbuHBhVPNCL5LU3d0d1bq6uqKaN7OUF3DxjnvChAlRrb29PapNnjzZPcaJEydGNW95S2/5x9SZnLxrO5RlFVMDOwTvUC/Gr0GMX4WRMH4x8gEAkCmaPAAAmaLJAwCQKZo8AACZoskDAJCphqbrvRSjlzpNTRJWJSBTp4D0eGlJ73hSj7FqHWBvbeKpU6cmbdPjXQvvGKumV/R4KdiBgYGo5p1jaqI39fUfSmLVk3o8QBXGr0GMX4WRMH5xJw8AQKZo8gAAZIomDwBApmjyAABkqqHBO48XevCmaxzK9IGpIY7U0EQ9z60Krnjnkzr1Zera1KkhlapQT9Wxp0i9Zh7vda0KKaW+1qnTVAJDwfh15m0yfhUu5vjFnTwAAJmiyQMAkCmaPAAAmaLJAwCQqYYG77xAgldramo65+1JfuCi3hmHUvbhqQpM1DNTlrfv1ADQUMI1qfv2eLNN1aPq9Uud4QuoF+PXIMavobmY4xejIQAAmaLJAwCQKZo8AACZoskDAJApYyYwAADyxJ08AACZoskDAJApmjwAAJmiyQMAkCmaPAAAmaLJAwCQKZo8AACZoskDAJApmjwAAJmiyQMAkCmaPAAAmaLJAwCQKZo8AACZoskDAJApmjwAAJmiyQMAkCmaPAAAmaLJAwCQKZo8AACZoskDAJApmjwAAJmiyQMAkCmaPAAAmfo/nwPfb79aHT0AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75fb0d2550>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"Original shape:\", X_train.shape)\n",
    "print(\"Normalized shape:\", X_train_normalized.shape)\n",
    "\n",
    "\n",
    "fig, axs = plt.subplots(1,2, figsize=(10, 3))\n",
    "axs = axs.ravel()\n",
    "\n",
    "axs[0].axis('off')\n",
    "axs[0].set_title('Train original')\n",
    "axs[0].imshow(X_train[0].squeeze(), cmap='gray')\n",
    "\n",
    "axs[1].axis('off')\n",
    "axs[1].set_title('Train normalized')\n",
    "axs[1].imshow(X_train_normalized[0].squeeze(), cmap='gray')\n",
    "\n",
    "fig, axs = plt.subplots(1,2, figsize=(10, 3))\n",
    "axs = axs.ravel()\n",
    "\n",
    "axs[0].axis('off')\n",
    "axs[0].set_title('Test original')\n",
    "axs[0].imshow(X_test[0].squeeze(), cmap='gray')\n",
    "\n",
    "axs[1].axis('off')\n",
    "axs[1].set_title('Test normalized')\n",
    "axs[1].imshow(X_test_normalized[0].squeeze(), cmap='gray')\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Original shape: (34799, 32, 32, 1)\n",
      "Translated shape: (32, 32, 1)\n",
      "Original shape: (12630, 32, 32, 1)\n",
      "Translated shape: (32, 32, 1)\n",
      "shape in/out: (32, 32, 1) (32, 32, 1)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75fb028f98>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75faf9ba20>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "X_train_scaling = random_scaling(X_train[0])\n",
    "X_test_scaling = random_scaling(X_test[0])\n",
    "\n",
    "print(\"Original shape:\", X_train.shape)\n",
    "print(\"Translated shape:\", X_train_scaling.shape)\n",
    "\n",
    "print(\"Original shape:\", X_test.shape)\n",
    "print(\"Translated shape:\", X_test_scaling.shape)\n",
    "\n",
    "fig, axs = plt.subplots(1,2, figsize=(10, 3))\n",
    "axs = axs.ravel()\n",
    "axs[0].axis('off')\n",
    "axs[0].set_title('Train original')\n",
    "axs[0].imshow(X_train[0].squeeze(), cmap='gray')\n",
    "\n",
    "axs[1].axis('off')\n",
    "axs[1].set_title('Train scaled')\n",
    "axs[1].imshow(X_train_scaling.squeeze(), cmap='gray')\n",
    "\n",
    "\n",
    "fig, axs = plt.subplots(1,2, figsize=(10, 3))\n",
    "\n",
    "axs[0].axis('off')\n",
    "axs[0].imshow(X_test[0].squeeze(), cmap='gray')\n",
    "axs[0].set_title('Test original')\n",
    "\n",
    "axs[1].axis('off')\n",
    "axs[1].imshow(X_test_scaling.squeeze(), cmap='gray')\n",
    "axs[1].set_title('Test scaled')\n",
    "\n",
    "print('shape in/out:', X_test[0].shape, X_test_scaling.shape)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Original shape: (34799, 32, 32, 1)\n",
      "Translated shape: (32, 32, 1)\n",
      "Test shape in/out: (12630, 32, 32, 1) (32, 32, 1)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75faef88d0>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75fae64470>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import cv2\n",
    "X_train_translate = random_translate(X_train[0])\n",
    "test_translate = random_translate(X_test[0])\n",
    "print(\"Original shape:\", X_train.shape)\n",
    "print(\"Translated shape:\", X_train_translate.shape)\n",
    "fig, axs = plt.subplots(1,2, figsize=(10, 3))\n",
    "axs = axs.ravel()\n",
    "axs[0].axis('off')\n",
    "axs[0].set_title('Train original')\n",
    "axs[0].imshow(X_train[0].squeeze(), cmap='gray')\n",
    "\n",
    "axs[1].axis('off')\n",
    "axs[1].set_title('Train translated')\n",
    "axs[1].imshow(X_train_translate.squeeze(), cmap='gray')\n",
    "\n",
    "fig, axs = plt.subplots(1,2, figsize=(10, 3))\n",
    "axs = axs.ravel()\n",
    "axs[0].axis('off')\n",
    "axs[0].set_title('Test original')\n",
    "axs[0].imshow(X_test[0].squeeze(), cmap='gray')\n",
    "\n",
    "axs[1].axis('off')\n",
    "axs[1].set_title('Test translated')\n",
    "axs[1].imshow(test_translate.squeeze(), cmap='gray')\n",
    "\n",
    "print('Test shape in/out:', X_test.shape, test_translate.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "shape in/out: (32, 32, 3) (32, 32, 1, 3)\n",
      "Test shape in/out: (12630, 32, 32, 1) (32, 32, 1)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75fadc2208>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75fad28f60>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_warped = random_warp(X_train[0])\n",
    "test_warped = random_warp(X_test[0])\n",
    "\n",
    "fig, axs = plt.subplots(1,2, figsize=(10, 3))\n",
    "\n",
    "axs[0].axis('off')\n",
    "axs[0].imshow(X_train[0].squeeze(), cmap='gray')\n",
    "axs[0].set_title('Train original')\n",
    "\n",
    "axs[1].axis('off')\n",
    "axs[1].imshow(train_warped.squeeze(), cmap='gray')\n",
    "axs[1].set_title('Train warped')\n",
    "\n",
    "print('shape in/out:', test_img.shape, test_dst.shape)\n",
    "\n",
    "fig, axs = plt.subplots(1,2, figsize=(10, 3))\n",
    "axs = axs.ravel()\n",
    "axs[0].axis('off')\n",
    "axs[0].set_title('Test original')\n",
    "axs[0].imshow(X_test[0].squeeze(), cmap='gray')\n",
    "\n",
    "axs[1].axis('off')\n",
    "axs[1].set_title('Test warped')\n",
    "axs[1].imshow(test_warped.squeeze(), cmap='gray')\n",
    "\n",
    "print('Test shape in/out:', X_test.shape, test_translate.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "shape in/out: (12630, 32, 32, 1) (32, 32, 1)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75fac066d8>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_brightness = random_brightness(X_train[0])\n",
    "test_brightness = random_brightness(X_test[0])\n",
    "fig, axs = plt.subplots(1,2, figsize=(10, 3))\n",
    "\n",
    "axs[0].axis('off')\n",
    "axs[0].imshow(X_train[0].squeeze(), cmap='gray')\n",
    "axs[0].set_title('Train original')\n",
    "\n",
    "axs[1].axis('off')\n",
    "axs[1].imshow(train_brightness.squeeze(), cmap='gray')\n",
    "axs[1].set_title('Train brightness adjusted')\n",
    "\n",
    "axs[0].axis('off')\n",
    "axs[0].imshow(X_test[0].squeeze(), cmap='gray')\n",
    "axs[0].set_title('Test original')\n",
    "\n",
    "axs[1].axis('off')\n",
    "axs[1].imshow(test_brightness.squeeze(), cmap='gray')\n",
    "axs[1].set_title('Test brightness adjusted')\n",
    "\n",
    "\n",
    "print('shape in/out:', X_test.shape, test_brightness.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model Architecture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [],
   "source": [
    "### Define your architecture here.\n",
    "### Feel free to use as many code cells as needed.\n",
    "### Define your architecture here.\n",
    "### Feel free to use as many code cells as needed.\n",
    "import tensorflow as tf\n",
    "\n",
    "from tensorflow.contrib.layers import flatten\n",
    "#from tensorflow.keras.layers import Flatten\n",
    "#from tf.keras.layers import Flatten\n",
    "#from tf.keras.layers import flatten\n",
    "\n",
    "EPOCHS = 60\n",
    "BATCH_SIZE = 128"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [],
   "source": [
    "### LeNet \n",
    "def LeNet(x, keep_prob):    \n",
    "    # Arguments used for tf.truncated_normal, randomly defines variables for the weights and biases for each layer\n",
    "    mu = 0\n",
    "    sigma = 0.1\n",
    "    \n",
    "    # Layer 1: Convolutional. Input = 32x32ximage_depth. Output = 28x28x6.\n",
    "    conv1_W = tf.Variable(tf.truncated_normal(shape=(5, 5, image_depth, 6), mean = mu, stddev = sigma))\n",
    "    conv1_b = tf.Variable(tf.zeros(6))\n",
    "    conv1   = tf.nn.conv2d(x, conv1_W, strides=[1, 1, 1, 1], padding='VALID', name='conv_1') + conv1_b\n",
    "    print (\"Convolution1: \", conv1)\n",
    "    \n",
    "    # Relu Activation\n",
    "    conv1 = tf.nn.relu(conv1)\n",
    "    \n",
    "    # Max Pooling. Input = 28x28x6. Output = 14x14x6.\n",
    "    conv1 = tf.nn.max_pool(conv1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')\n",
    "    \n",
    "    \n",
    "    # Layer 2: Convolutional. Output = 10x10x16.\n",
    "    conv2_W = tf.Variable(tf.truncated_normal(shape=(5, 5, 6, 16), mean = mu, stddev = sigma))\n",
    "    conv2_b = tf.Variable(tf.zeros(16))\n",
    "    conv2   = tf.nn.conv2d(conv1, conv2_W, strides=[1, 1, 1, 1], padding='VALID', name='conv_2') + conv2_b\n",
    "    print (\"Convolution2: \", conv2)\n",
    "    \n",
    "    # Relu Activation being done here\n",
    "    conv2 = tf.nn.relu(conv2)\n",
    "\n",
    "    # Max Pooling. Input = 10x10x16. Output = 5x5x16.\n",
    "    conv2 = tf.nn.max_pool(conv2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')\n",
    "           \n",
    "    # Flatten. Input = 5x5x16. Output = 400.\n",
    "    fc0   = flatten(conv2)\n",
    "    #fc0   = Flatten()(conv2)\n",
    "    \n",
    "    # Layer 3: Fully Connected. Input = 400. Output = 120.\n",
    "    fc1_W = tf.Variable(tf.truncated_normal(shape=(400, 120), mean = mu, stddev = sigma))\n",
    "    fc1_b = tf.Variable(tf.zeros(120))\n",
    "    fc1   = tf.matmul(fc0, fc1_W) + fc1_b\n",
    "    \n",
    "    # Relu Activation\n",
    "    fc1    = tf.nn.relu(fc1)\n",
    "    # Apply Dropout \n",
    "    fc1 = tf.nn.dropout(fc1, keep_prob)\n",
    "\n",
    "    # Layer 4: Fully Connected. Input = 120. Output = 84.\n",
    "    fc2_W  = tf.Variable(tf.truncated_normal(shape=(120, 84), mean = mu, stddev = sigma))\n",
    "    fc2_b  = tf.Variable(tf.zeros(84))\n",
    "    fc2    = tf.matmul(fc1, fc2_W) + fc2_b\n",
    "    \n",
    "    # Relu Activation\n",
    "    fc2    = tf.nn.relu(fc2)\n",
    "    # Apply Dropout \n",
    "    #fc2 = tf.nn.dropout(fc2, 0.75)\n",
    "\n",
    "    # Layer 5: Fully Connected. Input = 84. Output = 43.\n",
    "    fc3_W  = tf.Variable(tf.truncated_normal(shape=(84, n_classes), mean = mu, stddev = sigma))\n",
    "    fc3_b  = tf.Variable(tf.zeros(n_classes))\n",
    "    logits = tf.matmul(fc2, fc3_W) + fc3_b\n",
    "    \n",
    "    print (logits)\n",
    "    return logits"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train, Validate and Test the Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A validation set can be used to assess how well the model is performing. A low accuracy on the training and validation\n",
    "sets imply underfitting. A high accuracy on the training set but low accuracy on the validation set implies overfitting."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Convolution1:  Tensor(\"add_19:0\", shape=(?, 28, 28, 6), dtype=float32)\n",
      "Convolution2:  Tensor(\"add_20:0\", shape=(?, 10, 10, 16), dtype=float32)\n",
      "Tensor(\"add_23:0\", shape=(?, 43), dtype=float32)\n"
     ]
    }
   ],
   "source": [
    "### Train your model here.\n",
    "### Calculate and report the accuracy on the training and validation set.\n",
    "### Once a final model architecture is selected, \n",
    "### the accuracy on the test set should be calculated and reported as well.\n",
    "### Feel free to use as many code cells as needed.\n",
    "\n",
    "#import tensorflow.compat.v1 as tf\n",
    "#tf.disable_v2_behavior()\n",
    "\n",
    "# Create Placeholders for X and Y and One-Hot Encode the Labels\n",
    "x = tf.placeholder(tf.float32, (None, 32, 32, image_depth))\n",
    "y = tf.placeholder(tf.int32, (None))\n",
    "one_hot_y = tf.one_hot(y, n_classes)\n",
    "\n",
    "# dropout probability is the probability to keep units\n",
    "keep_prob = tf.placeholder(tf.float32)\n",
    "\n",
    "# Training Pipeline\n",
    "rate = 0.0009\n",
    "\n",
    "logits = LeNet(x, keep_prob)\n",
    "cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits = logits, labels = one_hot_y)\n",
    "loss_operation = tf.reduce_mean(cross_entropy)\n",
    "optimizer = tf.train.AdamOptimizer(learning_rate = rate)\n",
    "training_operation = optimizer.minimize(loss_operation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [
    {
     "ename": "ValueError",
     "evalue": "At least two variables have the same name: Variable_6",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-109-ac494c045310>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0mcorrect_prediction\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mequal\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0margmax\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlogits\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0margmax\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mone_hot_y\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0maccuracy_operation\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mreduce_mean\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcast\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcorrect_prediction\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfloat32\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 4\u001b[0;31m \u001b[0msaver\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrain\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mSaver\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      5\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mevaluate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX_data\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_data\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/lib/python3.6/site-packages/tensorflow/python/training/saver.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, var_list, reshape, sharded, max_to_keep, keep_checkpoint_every_n_hours, name, restore_sequentially, saver_def, builder, defer_build, allow_empty, write_version, pad_step_number, save_relative_paths, filename)\u001b[0m\n\u001b[1;32m   1138\u001b[0m     \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_filename\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfilename\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1139\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mdefer_build\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1140\u001b[0;31m       \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbuild\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1141\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msaver_def\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1142\u001b[0m       \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_check_saver_def\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/lib/python3.6/site-packages/tensorflow/python/training/saver.py\u001b[0m in \u001b[0;36mbuild\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m   1170\u001b[0m           \u001b[0mname\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_name\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1171\u001b[0m           \u001b[0mrestore_sequentially\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_restore_sequentially\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1172\u001b[0;31m           filename=self._filename)\n\u001b[0m\u001b[1;32m   1173\u001b[0m     \u001b[0;32melif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msaver_def\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_name\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1174\u001b[0m       \u001b[0;31m# Since self._name is used as a name_scope by builder(), we are\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/lib/python3.6/site-packages/tensorflow/python/training/saver.py\u001b[0m in \u001b[0;36mbuild\u001b[0;34m(self, names_to_saveables, reshape, sharded, max_to_keep, keep_checkpoint_every_n_hours, name, restore_sequentially, filename)\u001b[0m\n\u001b[1;32m    668\u001b[0m         \u001b[0munique\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    669\u001b[0m     \"\"\"\n\u001b[0;32m--> 670\u001b[0;31m     \u001b[0msaveables\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_ValidateAndSliceInputs\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnames_to_saveables\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    671\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mmax_to_keep\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    672\u001b[0m       \u001b[0mmax_to_keep\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/lib/python3.6/site-packages/tensorflow/python/training/saver.py\u001b[0m in \u001b[0;36m_ValidateAndSliceInputs\u001b[0;34m(self, names_to_saveables)\u001b[0m\n\u001b[1;32m    553\u001b[0m     \"\"\"\n\u001b[1;32m    554\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnames_to_saveables\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdict\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 555\u001b[0;31m       \u001b[0mnames_to_saveables\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mBaseSaverBuilder\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mOpListToDict\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnames_to_saveables\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    556\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    557\u001b[0m     \u001b[0msaveables\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/lib/python3.6/site-packages/tensorflow/python/training/saver.py\u001b[0m in \u001b[0;36mOpListToDict\u001b[0;34m(op_list)\u001b[0m\n\u001b[1;32m    531\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mname\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mnames_to_saveables\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    532\u001b[0m           raise ValueError(\"At least two variables have the same name: %s\" %\n\u001b[0;32m--> 533\u001b[0;31m                            name)\n\u001b[0m\u001b[1;32m    534\u001b[0m         \u001b[0mnames_to_saveables\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mname\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mvar\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    535\u001b[0m       \u001b[0;31m# pylint: enable=protected-access\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mValueError\u001b[0m: At least two variables have the same name: Variable_6"
     ]
    }
   ],
   "source": [
    "# Model Evaluation\n",
    "correct_prediction = tf.equal(tf.argmax(logits, 1), tf.argmax(one_hot_y, 1))\n",
    "accuracy_operation = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
    "saver = tf.train.Saver()\n",
    "\n",
    "def evaluate(X_data, y_data):\n",
    "    num_examples = len(X_data)\n",
    "    total_accuracy = 0\n",
    "    sess = tf.get_default_session()\n",
    "    for offset in range(0, num_examples, BATCH_SIZE):\n",
    "        batch_x, batch_y = X_data[offset:offset+BATCH_SIZE], y_data[offset:offset+BATCH_SIZE]\n",
    "        accuracy = sess.run(accuracy_operation, feed_dict={x: batch_x, y: batch_y, keep_prob: 1.0})\n",
    "        total_accuracy += (accuracy * len(batch_x))\n",
    "    return total_accuracy / num_examples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training...\n",
      "\n",
      "EPOCH 1 ...\n",
      "\n",
      "Training Accuracy = 0.123\n",
      "Validation Accuracy = 0.112\n",
      "\n",
      "EPOCH 2 ...\n",
      "\n",
      "Training Accuracy = 0.545\n",
      "Validation Accuracy = 0.513\n",
      "\n",
      "EPOCH 3 ...\n",
      "\n",
      "Training Accuracy = 0.709\n",
      "Validation Accuracy = 0.651\n",
      "\n",
      "EPOCH 4 ...\n",
      "\n",
      "Training Accuracy = 0.793\n",
      "Validation Accuracy = 0.729\n",
      "\n",
      "EPOCH 5 ...\n",
      "\n",
      "Training Accuracy = 0.833\n",
      "Validation Accuracy = 0.767\n",
      "\n",
      "EPOCH 6 ...\n",
      "\n",
      "Training Accuracy = 0.862\n",
      "Validation Accuracy = 0.774\n",
      "\n",
      "EPOCH 7 ...\n",
      "\n",
      "Training Accuracy = 0.881\n",
      "Validation Accuracy = 0.800\n",
      "\n",
      "EPOCH 8 ...\n",
      "\n",
      "Training Accuracy = 0.900\n",
      "Validation Accuracy = 0.828\n",
      "\n",
      "EPOCH 9 ...\n",
      "\n",
      "Training Accuracy = 0.903\n",
      "Validation Accuracy = 0.819\n",
      "\n",
      "EPOCH 10 ...\n",
      "\n",
      "Training Accuracy = 0.917\n",
      "Validation Accuracy = 0.836\n",
      "\n",
      "EPOCH 11 ...\n",
      "\n",
      "Training Accuracy = 0.920\n",
      "Validation Accuracy = 0.853\n",
      "\n",
      "EPOCH 12 ...\n",
      "\n",
      "Training Accuracy = 0.930\n",
      "Validation Accuracy = 0.840\n",
      "\n",
      "EPOCH 13 ...\n",
      "\n",
      "Training Accuracy = 0.935\n",
      "Validation Accuracy = 0.858\n",
      "\n",
      "EPOCH 14 ...\n",
      "\n",
      "Training Accuracy = 0.944\n",
      "Validation Accuracy = 0.865\n",
      "\n",
      "EPOCH 15 ...\n",
      "\n",
      "Training Accuracy = 0.951\n",
      "Validation Accuracy = 0.861\n",
      "\n",
      "EPOCH 16 ...\n",
      "\n",
      "Training Accuracy = 0.954\n",
      "Validation Accuracy = 0.872\n",
      "\n",
      "EPOCH 17 ...\n",
      "\n",
      "Training Accuracy = 0.956\n",
      "Validation Accuracy = 0.880\n",
      "\n",
      "EPOCH 18 ...\n",
      "\n",
      "Training Accuracy = 0.957\n",
      "Validation Accuracy = 0.867\n",
      "\n",
      "EPOCH 19 ...\n",
      "\n",
      "Training Accuracy = 0.961\n",
      "Validation Accuracy = 0.889\n",
      "\n",
      "EPOCH 20 ...\n",
      "\n",
      "Training Accuracy = 0.964\n",
      "Validation Accuracy = 0.882\n",
      "\n",
      "EPOCH 21 ...\n",
      "\n",
      "Training Accuracy = 0.965\n",
      "Validation Accuracy = 0.885\n",
      "\n",
      "EPOCH 22 ...\n",
      "\n",
      "Training Accuracy = 0.969\n",
      "Validation Accuracy = 0.889\n",
      "\n",
      "EPOCH 23 ...\n",
      "\n",
      "Training Accuracy = 0.971\n",
      "Validation Accuracy = 0.890\n",
      "\n",
      "EPOCH 24 ...\n",
      "\n",
      "Training Accuracy = 0.972\n",
      "Validation Accuracy = 0.894\n",
      "\n",
      "EPOCH 25 ...\n",
      "\n",
      "Training Accuracy = 0.974\n",
      "Validation Accuracy = 0.892\n",
      "\n",
      "EPOCH 26 ...\n",
      "\n",
      "Training Accuracy = 0.976\n",
      "Validation Accuracy = 0.887\n",
      "\n",
      "EPOCH 27 ...\n",
      "\n",
      "Training Accuracy = 0.978\n",
      "Validation Accuracy = 0.897\n",
      "\n",
      "EPOCH 28 ...\n",
      "\n",
      "Training Accuracy = 0.978\n",
      "Validation Accuracy = 0.897\n",
      "\n",
      "EPOCH 29 ...\n",
      "\n",
      "Training Accuracy = 0.980\n",
      "Validation Accuracy = 0.903\n",
      "\n",
      "EPOCH 30 ...\n",
      "\n",
      "Training Accuracy = 0.980\n",
      "Validation Accuracy = 0.909\n",
      "\n",
      "EPOCH 31 ...\n",
      "\n",
      "Training Accuracy = 0.982\n",
      "Validation Accuracy = 0.903\n",
      "\n",
      "EPOCH 32 ...\n",
      "\n",
      "Training Accuracy = 0.979\n",
      "Validation Accuracy = 0.904\n",
      "\n",
      "EPOCH 33 ...\n",
      "\n",
      "Training Accuracy = 0.982\n",
      "Validation Accuracy = 0.898\n",
      "\n",
      "EPOCH 34 ...\n",
      "\n",
      "Training Accuracy = 0.982\n",
      "Validation Accuracy = 0.906\n",
      "\n",
      "EPOCH 35 ...\n",
      "\n",
      "Training Accuracy = 0.986\n",
      "Validation Accuracy = 0.914\n",
      "\n",
      "EPOCH 36 ...\n",
      "\n",
      "Training Accuracy = 0.985\n",
      "Validation Accuracy = 0.914\n",
      "\n",
      "EPOCH 37 ...\n",
      "\n",
      "Training Accuracy = 0.985\n",
      "Validation Accuracy = 0.910\n",
      "\n",
      "EPOCH 38 ...\n",
      "\n",
      "Training Accuracy = 0.984\n",
      "Validation Accuracy = 0.920\n",
      "\n",
      "EPOCH 39 ...\n",
      "\n",
      "Training Accuracy = 0.985\n",
      "Validation Accuracy = 0.909\n",
      "\n",
      "EPOCH 40 ...\n",
      "\n",
      "Training Accuracy = 0.986\n",
      "Validation Accuracy = 0.909\n",
      "\n",
      "EPOCH 41 ...\n",
      "\n",
      "Training Accuracy = 0.988\n",
      "Validation Accuracy = 0.909\n",
      "\n",
      "EPOCH 42 ...\n",
      "\n",
      "Training Accuracy = 0.988\n",
      "Validation Accuracy = 0.912\n",
      "\n",
      "EPOCH 43 ...\n",
      "\n",
      "Training Accuracy = 0.989\n",
      "Validation Accuracy = 0.917\n",
      "\n",
      "EPOCH 44 ...\n",
      "\n",
      "Training Accuracy = 0.988\n",
      "Validation Accuracy = 0.918\n",
      "\n",
      "EPOCH 45 ...\n",
      "\n",
      "Training Accuracy = 0.989\n",
      "Validation Accuracy = 0.912\n",
      "\n",
      "EPOCH 46 ...\n",
      "\n",
      "Training Accuracy = 0.987\n",
      "Validation Accuracy = 0.921\n",
      "\n",
      "EPOCH 47 ...\n",
      "\n",
      "Training Accuracy = 0.990\n",
      "Validation Accuracy = 0.924\n",
      "\n",
      "EPOCH 48 ...\n",
      "\n",
      "Training Accuracy = 0.990\n",
      "Validation Accuracy = 0.911\n",
      "\n",
      "EPOCH 49 ...\n",
      "\n",
      "Training Accuracy = 0.990\n",
      "Validation Accuracy = 0.927\n",
      "\n",
      "EPOCH 50 ...\n",
      "\n",
      "Training Accuracy = 0.991\n",
      "Validation Accuracy = 0.921\n",
      "\n",
      "EPOCH 51 ...\n",
      "\n",
      "Training Accuracy = 0.991\n",
      "Validation Accuracy = 0.922\n",
      "\n",
      "EPOCH 52 ...\n",
      "\n",
      "Training Accuracy = 0.991\n",
      "Validation Accuracy = 0.933\n",
      "\n",
      "EPOCH 53 ...\n",
      "\n",
      "Training Accuracy = 0.992\n",
      "Validation Accuracy = 0.924\n",
      "\n",
      "EPOCH 54 ...\n",
      "\n",
      "Training Accuracy = 0.992\n",
      "Validation Accuracy = 0.927\n",
      "\n",
      "EPOCH 55 ...\n",
      "\n",
      "Training Accuracy = 0.991\n",
      "Validation Accuracy = 0.924\n",
      "\n",
      "EPOCH 56 ...\n",
      "\n",
      "Training Accuracy = 0.992\n",
      "Validation Accuracy = 0.932\n",
      "\n",
      "EPOCH 57 ...\n",
      "\n",
      "Training Accuracy = 0.992\n",
      "Validation Accuracy = 0.926\n",
      "\n",
      "EPOCH 58 ...\n",
      "\n",
      "Training Accuracy = 0.991\n",
      "Validation Accuracy = 0.924\n",
      "\n",
      "EPOCH 59 ...\n",
      "\n",
      "Training Accuracy = 0.993\n",
      "Validation Accuracy = 0.926\n",
      "\n",
      "EPOCH 60 ...\n",
      "\n",
      "Training Accuracy = 0.994\n",
      "Validation Accuracy = 0.929\n",
      "\n",
      "Model saved\n"
     ]
    }
   ],
   "source": [
    "### Train your model here.\n",
    "\n",
    "with tf.Session() as sess:\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "    num_examples = len(X_train)\n",
    "    \n",
    "    print(\"Training...\")\n",
    "    print()\n",
    "    for i in range(EPOCHS):\n",
    "        X_train, y_train = shuffle(X_train, y_train)\n",
    "        for offset in range(0, num_examples, BATCH_SIZE):\n",
    "            end = offset + BATCH_SIZE\n",
    "            batch_x, batch_y = X_train[offset:end], y_train[offset:end]\n",
    "            sess.run(training_operation, feed_dict={x: batch_x, y: batch_y, keep_prob: 0.5})\n",
    "        \n",
    "        # Calculate and report the validation accuracy \n",
    "        training_accuracy = evaluate(X_train, y_train)\n",
    "        validation_accuracy = evaluate(X_valid, y_valid)\n",
    "\n",
    "        print(\"EPOCH {} ...\".format(i+1))\n",
    "        print()   \n",
    "        print(\"Training Accuracy = {:.3f}\".format(training_accuracy))\n",
    "     \n",
    "        #print(\"EPOCH {} ...\".format(i+1))\n",
    "        print(\"Validation Accuracy = {:.3f}\".format(validation_accuracy))\n",
    "        print()\n",
    "        \n",
    "    # Save the model    \n",
    "    saver.save(sess, './traffic_signs')\n",
    "    print(\"Model saved\")\n",
    "\n",
    "### Calculate and report the accuracy on the training and validation set.\n",
    "### Once a final model architecture is selected, \n",
    "### the accuracy on the test set should be calculated and reported as well.\n",
    "### Feel free to use as many code cells as needed."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 2.5: Test a Model on New Images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from ./traffic_signs\n",
      "Test Set Accuracy = 0.050\n"
     ]
    }
   ],
   "source": [
    "# Now (drumroll) evaluate the accuracy of the model on the test dataset\n",
    "\n",
    "with tf.Session() as sess:\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "    saver2 = tf.train.import_meta_graph('./traffic_signs.meta')\n",
    "    saver2.restore(sess, \"./traffic_signs\")\n",
    "    test_accuracy = evaluate(X_test, y_test)\n",
    "    print(\"Test Set Accuracy = {:.3f}\".format(test_accuracy))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "\n",
    "## Step 3: Test a Model on New Images\n",
    "\n",
    "To give yourself more insight into how your model is working, download at least five pictures of German traffic signs from the web and use your model to predict the traffic sign type.\n",
    "\n",
    "You may find `signnames.csv` useful as it contains mappings from the class id (integer) to the actual sign name."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load and Output the Images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "14\n",
      "1\n",
      "25\n",
      "9\n",
      "5\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75fa7a29b0>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75fa353518>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75fa028908>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75e4cab588>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75e4bf64a8>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "### Load the images and plot them here.\n",
    "### Feel free to use as many code cells as needed.\n",
    "\n",
    "### Load the images and plot them here.\n",
    "### Load and output the web images\n",
    "from scipy import misc\n",
    "import imageio\n",
    "\n",
    "# Create list with downloaded images\n",
    "image_features = []\n",
    "for i in range(1,6):\n",
    "    # Save image data as ndarrays in list\n",
    "    #image_features.append(misc.imread(parent_dir + \"/traffic-signs-data/web_images/\" + str(i) + \".jpg\", mode='RGB'))\n",
    "    image_features.append(imageio.imread(\"./traffic-signs-data/online_files/\" + str(i) + \".jpg\"))\n",
    "    \n",
    "# Create ndarrays with image_features and labels list\n",
    "X_online_test = np.array(image_features)\n",
    "Y_online_test = np.array([14, 1, 25, 9, 5])\n",
    "\n",
    "# Check that the same amount of features and labels was stored\n",
    "assert(len(X_online_test) == len(Y_online_test))\n",
    "\n",
    "# Print out all images and the respective labels\n",
    "for i, im in enumerate (X_online_test):\n",
    "    print (Y_online_test[i])\n",
    "    image = im.squeeze()\n",
    "    plt.figure(figsize=(2,2))\n",
    "    plt.imshow(image, cmap=\"gray\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predict the Sign Type for Each Image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5, 32, 32, 3)\n",
      "5 32 32 1\n",
      "Restoring the model . . . \n",
      "INFO:tensorflow:Restoring parameters from ./traffic_signs\n",
      "Model restored.\n",
      "Samples:  5\n",
      "label :  [ 7 12  8 30  8]\n",
      "Prediction :  [False False False False False]\n",
      "Probability :  [[-0.05272397  0.0200924  -0.02898298  0.01673755 -0.08419205 -0.03762679\n",
      "   0.03831628  0.06204474  0.04633746 -0.00369399  0.00028991 -0.03284062\n",
      "   0.04292487  0.00033031 -0.07380268  0.03232404  0.0106379  -0.02614592\n",
      "   0.05854928 -0.06415991 -0.02431366  0.05885797  0.01719164  0.02671637\n",
      "   0.00368808 -0.06297641 -0.01792141 -0.03138649 -0.04680558  0.04311337\n",
      "   0.04592644 -0.00433451 -0.03371209  0.01040761 -0.07567389  0.00718888\n",
      "   0.01283679 -0.01119118 -0.05361966  0.02946863  0.00955878  0.05358161\n",
      "   0.01561081]\n",
      " [-0.08777672  0.04671842 -0.03775029  0.0331738  -0.11082038 -0.05027482\n",
      "   0.08584332  0.06615224  0.07362001  0.00471378 -0.00890311 -0.03574736\n",
      "   0.08802235  0.021572   -0.10240909  0.06745846  0.02125342 -0.03268972\n",
      "   0.06655553 -0.06983723 -0.0063286   0.0763821  -0.01714872  0.03738204\n",
      "  -0.00049521 -0.09494743 -0.02475868 -0.06174196 -0.07132331  0.05949117\n",
      "   0.07527985 -0.00752712 -0.05320283  0.0195926  -0.11854727  0.01869976\n",
      "  -0.00885749 -0.00057224 -0.06994482  0.0433789   0.00977214  0.0757581\n",
      "   0.03878306]\n",
      " [-0.09376074  0.0560858  -0.03345569  0.0245378  -0.10693439 -0.05247346\n",
      "   0.08020743  0.05258867  0.08438073 -0.00241724  0.00581015 -0.03225713\n",
      "   0.06919475  0.02389145 -0.09105806  0.0638646   0.04512401 -0.04085054\n",
      "   0.04898861 -0.05798281 -0.00939433  0.06057328 -0.01228815  0.04343882\n",
      "  -0.00605242 -0.078105   -0.0126108  -0.06189059 -0.0695322   0.05294734\n",
      "   0.08094086 -0.00869423 -0.06637179  0.01607309 -0.1014633   0.03004041\n",
      "  -0.00274768 -0.00064941 -0.05030634  0.021656    0.0078701   0.06637596\n",
      "   0.04310768]\n",
      " [-0.06790014  0.05463304 -0.03993287  0.01449109 -0.10981572 -0.04017065\n",
      "   0.06604658  0.06630396  0.06010216  0.00708817  0.00284931 -0.02451229\n",
      "   0.06885841  0.030113   -0.07616169  0.04884559  0.03089231 -0.02948997\n",
      "   0.04251918 -0.05742372 -0.02047449  0.06624461 -0.01597602  0.05114175\n",
      "  -0.00209544 -0.08034035 -0.02445792 -0.05671862 -0.0660364   0.04063787\n",
      "   0.07529487 -0.00217792 -0.07658137  0.01589962 -0.09525847  0.00281279\n",
      "  -0.011819    0.00988896 -0.05837277  0.01057923  0.01024974  0.07154107\n",
      "   0.03972166]\n",
      " [-0.0990802   0.06089159 -0.03384258  0.0196156  -0.11072821 -0.0348698\n",
      "   0.07570217  0.0534768   0.07912411  0.01047641 -0.00279518 -0.03494272\n",
      "   0.07616524  0.03378806 -0.10314354  0.07545938  0.04195642 -0.04075868\n",
      "   0.05139469 -0.05371378 -0.01369004  0.06562958 -0.00595029  0.03812341\n",
      "  -0.00747221 -0.10210795 -0.02265135 -0.07147704 -0.06164541  0.05473521\n",
      "   0.07781244 -0.00700341 -0.05928572  0.03739    -0.11872007  0.02254168\n",
      "  -0.01066274 -0.01003572 -0.05917569  0.02621492  0.00333758  0.06409727\n",
      "   0.05332711]]\n",
      "Labels :  [ 7 12  8 30  8]\n"
     ]
    }
   ],
   "source": [
    "### Feel free to use as many code cells as needed.\n",
    "# Covert to Grayscale & Normalize\n",
    "print(X_online_test.shape)\n",
    "\n",
    "def conv_rgb2gray(rgb):\n",
    "    #conv_rgb2gray\n",
    "    return np.dot(rgb, [0.299, 0.587, 0.114])\n",
    "\n",
    "if (X_online_test.shape[3] == 3):\n",
    "    # Convert to Grayscale\n",
    "    X_online_test_gray = conv_rgb2gray(X_online_test)\n",
    "    \n",
    "    # Normalize Grayscale Images\n",
    "    from sklearn import preprocessing\n",
    "    \n",
    "    for i, picture in enumerate(X_online_test_gray):\n",
    "        X_online_test_gray[i] = preprocessing.normalize(picture, norm='l2', axis=1, copy=True, return_norm=False)\n",
    "    \n",
    "    # Reshape Grayscale Pictures (Add Dimension 1)\n",
    "    X_online_test_norm = X_online_test_gray.reshape(X_online_test.shape[0], X_online_test.shape[1], X_online_test.shape[2], 1)\n",
    "    print(X_online_test.shape[0], X_online_test.shape[1], X_online_test.shape[2], 1)\n",
    "    X_online_test = X_online_test_norm\n",
    "    #X_online_test = X_online_test.astype(np.float32)\n",
    "    \n",
    "    # Save Variable Image Depth\n",
    "    image_depth = X_test.shape[3]\n",
    "\n",
    "\n",
    "### Run the predictions here and use the model to output the prediction for each image.\n",
    "### Make sure to pre-process the images with the same pre-processing pipeline used earlier.\n",
    "\n",
    "def prediction(X_data, y_data):\n",
    "    num_examples = int(len(X_data))\n",
    "    print (\"Samples: \", num_examples)\n",
    "    sess = tf.get_default_session()\n",
    "    \"\"\"\n",
    "    #for offset in range(0, num_examples, BATCH_SIZE):\n",
    "    #for offset in range(0, num_examples):\n",
    "    #batch_x, batch_y = X_data[offset:offset+BATCH_SIZE], y_data[offset:offset+BATCH_SIZE]\n",
    "    batch_x, batch_y = X_data, y_data\n",
    "    \"\"\"\n",
    "    for offset in range(0, num_examples, BATCH_SIZE):\n",
    "    #for offset in range(0, num_examples):\n",
    "        batch_x, batch_y = X_data[offset:offset+BATCH_SIZE], y_data[offset:offset+BATCH_SIZE]\n",
    "    #batch_x, batch_y = X_data, y_data\n",
    "    with tf.Session() as sess:\n",
    "        sess.run(tf.global_variables_initializer())\n",
    "        predict = sess.run(correct_prediction, feed_dict={x: batch_x, y: batch_y, keep_prob: 1.0})\n",
    "        #print (\"prediction : \", predict)\n",
    "        label = sess.run(tf.argmax(logits, 1), feed_dict={x: batch_x, y: batch_y, keep_prob: 1.0})\n",
    "        probabilities = sess.run(logits, feed_dict={x: batch_x, y: batch_y, keep_prob: 1.0})\n",
    "        print (\"label : \", label)\n",
    "        #print (\"probabilities : \", probabilities)\n",
    "    return predict, probabilities, label\n",
    "\n",
    "# Open session with restored model\n",
    "with tf.Session() as sess:\n",
    "    #sess.run(tf.global_variables_initializer())\n",
    "    # Restore model\n",
    "    print(\"Restoring the model . . . \")\n",
    "    saver.restore(sess, './traffic_signs')\n",
    "    print(\"Model restored.\")\n",
    "    \n",
    "    # Verify accuracy of the trained model via test data\n",
    "    pred, prob, label = prediction(X_online_test, Y_online_test)\n",
    "    print(\"Prediction : \", pred)\n",
    "    print(\"Probability : \", prob)\n",
    "    print(\"Labels : \", label)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Analyze Performance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Testing...\n",
      "EPOCH 60 ...\n",
      "Test Accuracy = 0.007\n",
      "\n",
      "Model saved\n"
     ]
    }
   ],
   "source": [
    "### Train your model here.\n",
    "\n",
    "### Calculate and report the accuracy on the training and validation set.\n",
    "### Once a final model architecture is selected, \n",
    "### the accuracy on the test set should be calculated and reported as well.\n",
    "### Feel free to use as many code cells as needed.\n",
    "\n",
    "### Validate your model here.\n",
    "rate=0.0009\n",
    "\n",
    "with tf.Session() as sess:\n",
    "    sess.run(tf.global_variables_initializer())   \n",
    "    print(\"Testing...\")\n",
    "    for i in range(EPOCHS):\n",
    "        #X_test, y_test = shuffle(X_test, y_test)\n",
    "        \n",
    "        # Calculate and report the validation accuracy \n",
    "        validation_accuracy = evaluate(X_test, y_test)\n",
    "    print(\"EPOCH {} ...\".format(i+1))\n",
    "    print(\"Test Accuracy = {:.3f}\".format(validation_accuracy))\n",
    "    print()       \n",
    "        \n",
    "    # Save the model    \n",
    "    saver.save(sess, './traffic_signs')\n",
    "    print(\"Model saved\")\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Output Top 5 Softmax Probabilities For Each Image Found on the Web"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For each of the new images, print out the model's softmax probabilities to show the **certainty** of the model's predictions (limit the output to the top 5 probabilities for each image). [`tf.nn.top_k`](https://www.tensorflow.org/versions/r0.12/api_docs/python/nn.html#top_k) could prove helpful here. \n",
    "\n",
    "The example below demonstrates how tf.nn.top_k can be used to find the top k predictions for each image.\n",
    "\n",
    "`tf.nn.top_k` will return the values and indices (class ids) of the top k predictions. So if k=3, for each sign, it'll return the 3 largest probabilities (out of a possible 43) and the correspoding class ids.\n",
    "\n",
    "Take this numpy array as an example. The values in the array represent predictions. The array contains softmax probabilities for five candidate images with six possible classes. `tf.nn.top_k` is used to choose the three classes with the highest probability:\n",
    "\n",
    "```\n",
    "# (5, 6) array\n",
    "a = np.array([[ 0.24879643,  0.07032244,  0.12641572,  0.34763842,  0.07893497,\n",
    "         0.12789202],\n",
    "       [ 0.28086119,  0.27569815,  0.08594638,  0.0178669 ,  0.18063401,\n",
    "         0.15899337],\n",
    "       [ 0.26076848,  0.23664738,  0.08020603,  0.07001922,  0.1134371 ,\n",
    "         0.23892179],\n",
    "       [ 0.11943333,  0.29198961,  0.02605103,  0.26234032,  0.1351348 ,\n",
    "         0.16505091],\n",
    "       [ 0.09561176,  0.34396535,  0.0643941 ,  0.16240774,  0.24206137,\n",
    "         0.09155967]])\n",
    "```\n",
    "\n",
    "Running it through `sess.run(tf.nn.top_k(tf.constant(a), k=3))` produces:\n",
    "\n",
    "```\n",
    "TopKV2(values=array([[ 0.34763842,  0.24879643,  0.12789202],\n",
    "       [ 0.28086119,  0.27569815,  0.18063401],\n",
    "       [ 0.26076848,  0.23892179,  0.23664738],\n",
    "       [ 0.29198961,  0.26234032,  0.16505091],\n",
    "       [ 0.34396535,  0.24206137,  0.16240774]]), indices=array([[3, 0, 5],\n",
    "       [0, 1, 4],\n",
    "       [0, 5, 1],\n",
    "       [1, 3, 5],\n",
    "       [1, 4, 3]], dtype=int32))\n",
    "```\n",
    "\n",
    "Looking just at the first row we get `[ 0.34763842,  0.24879643,  0.12789202]`, you can confirm these are the 3 largest probabilities in `a`. You'll also notice `[3, 0, 5]` are the corresponding indices."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "top 5 :  TopKV2(values=array([[ 0.06204474,  0.05885797,  0.05854928],\n",
      "       [ 0.08802235,  0.08584332,  0.0763821 ],\n",
      "       [ 0.08438073,  0.08094086,  0.08020743],\n",
      "       [ 0.07529487,  0.07154107,  0.06885841],\n",
      "       [ 0.07912411,  0.07781244,  0.07616524]], dtype=float32), indices=array([[ 7, 21, 18],\n",
      "       [12,  6, 21],\n",
      "       [ 8, 30,  6],\n",
      "       [30, 41, 12],\n",
      "       [ 8, 30, 12]], dtype=int32))\n"
     ]
    }
   ],
   "source": [
    "### Print out the top five softmax probabilities for the predictions on the German traffic sign images found on the web. \n",
    "### Feel free to use as many code cells as needed.\n",
    "with tf.Session() as sess:\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "    top_5 = sess.run(tf.nn.top_k(tf.constant(prob), k=3))\n",
    "print(\"top 5 : \", top_5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'\\nprint(len(X_select_rgb))\\n\\nfor image in select_images:\\n    X_select_rgb = np.sum(image, axis=0, keepdims=True)\\n    \\nselect_images_normalized = (X_select_rgb - 128)/128\\n\\nprint(X_select_rgb.shape)\\n'"
      ]
     },
     "execution_count": 115,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f75fad8cc18>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "### Load the images and plot them here.\n",
    "### Feel free to use as many code cells as needed.\n",
    "\n",
    "#reading in an image\n",
    "import os\n",
    "import matplotlib.image as mpimg\n",
    "import cv2\n",
    "\n",
    "fig, axs = plt.subplots(2,4, figsize=(4, 2))\n",
    "fig.subplots_adjust(hspace = .2, wspace=.001)\n",
    "axs = axs.ravel()\n",
    "\n",
    "select_images = []\n",
    "select_images_gray = []\n",
    "\n",
    "def conv_rgb2gray(rgb):\n",
    "    return np.dot(rgb, [0.299, 0.587, 0.114])\n",
    "\n",
    "for i, img in enumerate(os.listdir('traffic-signs-data/online_files/')):\n",
    "    image = cv2.imread('traffic-signs-data/online_files/' + img)\n",
    "    #print(len(image))\n",
    "    #axs[i].axis('off')\n",
    "    #axs[i].imshow(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))\n",
    "    select_images.append(image)\n",
    "    #print(\"-----------------------\")\n",
    "    #select_images_gray.append(conv_rgb2gray(image))\n",
    "\n",
    "# Convert to grayscale\n",
    "select_images = np.asarray(select_images)\n",
    "X_select_rgb = select_images\n",
    "#X_select_gry = np.sum(X_select_rgb/3, axis=3, keepdims=True)\n",
    "\n",
    "\"\"\"\n",
    "print(len(X_select_rgb))\n",
    "\n",
    "for image in select_images:\n",
    "    X_select_rgb = np.sum(image, axis=0, keepdims=True)\n",
    "    \n",
    "select_images_normalized = (X_select_rgb - 128)/128\n",
    "\n",
    "print(X_select_rgb.shape)\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Additional work needed getting errors and will continue to work on this"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from ./traffic_signs\n",
      "6\n",
      "5\n"
     ]
    },
    {
     "ename": "ValueError",
     "evalue": "Cannot feed value of shape (1, 32, 3) for Tensor 'Placeholder_17:0', which has shape '(?, 32, 32, 1)'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-116-3fa232055827>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     11\u001b[0m     \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX_select_rgb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     12\u001b[0m     \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mselect_labels\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 13\u001b[0;31m     \u001b[0mselect_accuracy\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mevaluate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mselect_images_normalized\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mselect_labels\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     14\u001b[0m     \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Test Set Accuracy = {:.3f}\"\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mformat\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mselect_accuracy\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     15\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Done\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-27-ac494c045310>\u001b[0m in \u001b[0;36mevaluate\u001b[0;34m(X_data, y_data)\u001b[0m\n\u001b[1;32m     10\u001b[0m     \u001b[0;32mfor\u001b[0m \u001b[0moffset\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnum_examples\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mBATCH_SIZE\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     11\u001b[0m         \u001b[0mbatch_x\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbatch_y\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mX_data\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0moffset\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0moffset\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0mBATCH_SIZE\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_data\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0moffset\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0moffset\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0mBATCH_SIZE\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 12\u001b[0;31m         \u001b[0maccuracy\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msess\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0maccuracy_operation\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m{\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mbatch_x\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mbatch_y\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkeep_prob\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;36m1.0\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     13\u001b[0m         \u001b[0mtotal_accuracy\u001b[0m \u001b[0;34m+=\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0maccuracy\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mbatch_x\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     14\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0mtotal_accuracy\u001b[0m \u001b[0;34m/\u001b[0m \u001b[0mnum_examples\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m    893\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    894\u001b[0m       result = self._run(None, fetches, feed_dict, options_ptr,\n\u001b[0;32m--> 895\u001b[0;31m                          run_metadata_ptr)\n\u001b[0m\u001b[1;32m    896\u001b[0m       \u001b[0;32mif\u001b[0m \u001b[0mrun_metadata\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    897\u001b[0m         \u001b[0mproto_data\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtf_session\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mTF_GetBuffer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrun_metadata_ptr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_run\u001b[0;34m(self, handle, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m   1098\u001b[0m                 \u001b[0;34m'Cannot feed value of shape %r for Tensor %r, '\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1099\u001b[0m                 \u001b[0;34m'which has shape %r'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1100\u001b[0;31m                 % (np_val.shape, subfeed_t.name, str(subfeed_t.get_shape())))\n\u001b[0m\u001b[1;32m   1101\u001b[0m           \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mgraph\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mis_feedable\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msubfeed_t\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1102\u001b[0m             \u001b[0;32mraise\u001b[0m \u001b[0mValueError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'Tensor %s may not be fed.'\u001b[0m \u001b[0;34m%\u001b[0m \u001b[0msubfeed_t\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mValueError\u001b[0m: Cannot feed value of shape (1, 32, 3) for Tensor 'Placeholder_17:0', which has shape '(?, 32, 32, 1)'"
     ]
    }
   ],
   "source": [
    "### Run the predictions here.\n",
    "### Feel free to use as many code cells as needed.\n",
    "### Reviewer - Please pay additional attention here;  This is throwing errors\n",
    "#select_labels = [11, 3, 1, 12, 38, 34, 18, 25]\n",
    "select_labels = [14, 1, 25, 9, 5]\n",
    "\n",
    "with tf.Session() as sess:\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "    saver3 = tf.train.import_meta_graph('./traffic_signs.meta')\n",
    "    saver3.restore(sess, \"./traffic_signs\")\n",
    "    print(len(X_select_rgb))\n",
    "    print(len(select_labels))\n",
    "    select_accuracy = evaluate(select_images_normalized, select_labels)\n",
    "    print(\"Test Set Accuracy = {:.3f}\".format(select_accuracy))\n",
    "print(\"Done\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Project Writeup\n",
    "\n",
    "Once you have completed the code implementation, document your results in a project writeup using this [template](https://github.com/udacity/CarND-Traffic-Sign-Classifier-Project/blob/master/writeup_template.md) as a guide. The writeup can be in a markdown or pdf file. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> **Note**: Once you have completed all of the code implementations and successfully answered each question above, you may finalize your work by exporting the iPython Notebook as an HTML document. You can do this by using the menu above and navigating to  \\n\",\n",
    "    \"**File -> Download as -> HTML (.html)**. Include the finished document along with this notebook as your submission."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "\n",
    "## Step 4 (Optional): Visualize the Neural Network's State with Test Images\n",
    "\n",
    " This Section is not required to complete but acts as an additional excersise for understaning the output of a neural network's weights. While neural networks can be a great learning device they are often referred to as a black box. We can understand what the weights of a neural network look like better by plotting their feature maps. After successfully training your neural network you can see what it's feature maps look like by plotting the output of the network's weight layers in response to a test stimuli image. From these plotted feature maps, it's possible to see what characteristics of an image the network finds interesting. For a sign, maybe the inner network feature maps react with high activation to the sign's boundary outline or to the contrast in the sign's painted symbol.\n",
    "\n",
    " Provided for you below is the function code that allows you to get the visualization output of any tensorflow weight layer you want. The inputs to the function should be a stimuli image, one used during training or a new one you provided, and then the tensorflow variable name that represents the layer's state during the training process, for instance if you wanted to see what the [LeNet lab's](https://classroom.udacity.com/nanodegrees/nd013/parts/fbf77062-5703-404e-b60c-95b78b2f3f9e/modules/6df7ae49-c61c-4bb2-a23e-6527e69209ec/lessons/601ae704-1035-4287-8b11-e2c2716217ad/concepts/d4aca031-508f-4e0b-b493-e7b706120f81) feature maps looked like for it's second convolutional layer you could enter conv2 as the tf_activation variable.\n",
    "\n",
    "For an example of what feature map outputs look like, check out NVIDIA's results in their paper [End-to-End Deep Learning for Self-Driving Cars](https://devblogs.nvidia.com/parallelforall/deep-learning-self-driving-cars/) in the section Visualization of internal CNN State. NVIDIA was able to show that their network's inner weights had high activations to road boundary lines by comparing feature maps from an image with a clear path to one without. Try experimenting with a similar test to show that your trained network's weights are looking for interesting features, whether it's looking at differences in feature maps from images with or without a sign, or even what feature maps look like in a trained network vs a completely untrained one on the same sign image.\n",
    "\n",
    "<figure>\n",
    " <img src=\"visualize_cnn.png\" width=\"380\" alt=\"Combined Image\" />\n",
    " <figcaption>\n",
    " <p></p> \n",
    " <p style=\"text-align: center;\"> Your output should look something like this (above)</p> \n",
    " </figcaption>\n",
    "</figure>\n",
    " <p></p> \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "### Visualize your network's feature maps here.\n",
    "### Feel free to use as many code cells as needed.\n",
    "\n",
    "# image_input: the test image being fed into the network to produce the feature maps\n",
    "# tf_activation: should be a tf variable name used during your training procedure that represents the calculated state of a specific weight layer\n",
    "# activation_min/max: can be used to view the activation contrast in more detail, by default matplot sets min and max to the actual min and max values of the output\n",
    "# plt_num: used to plot out multiple different weight feature map sets on the same block, just extend the plt number for each new feature map entry\n",
    "\n",
    "def outputFeatureMap(image_input, tf_activation, activation_min=-1, activation_max=-1 ,plt_num=1):\n",
    "    # Here make sure to preprocess your image_input in a way your network expects\n",
    "    # with size, normalization, ect if needed\n",
    "    # image_input =\n",
    "    # Note: x should be the same name as your network's tensorflow data placeholder variable\n",
    "    # If you get an error tf_activation is not defined it may be having trouble accessing the variable from inside a function\n",
    "    activation = tf_activation.eval(session=sess,feed_dict={x : image_input})\n",
    "    featuremaps = activation.shape[3]\n",
    "    plt.figure(plt_num, figsize=(15,15))\n",
    "    for featuremap in range(featuremaps):\n",
    "        plt.subplot(6,8, featuremap+1) # sets the number of feature maps to show on each row and column\n",
    "        plt.title('FeatureMap ' + str(featuremap)) # displays the feature map number\n",
    "        if activation_min != -1 & activation_max != -1:\n",
    "            plt.imshow(activation[0,:,:, featuremap], interpolation=\"nearest\", vmin =activation_min, vmax=activation_max, cmap=\"gray\")\n",
    "        elif activation_max != -1:\n",
    "            plt.imshow(activation[0,:,:, featuremap], interpolation=\"nearest\", vmax=activation_max, cmap=\"gray\")\n",
    "        elif activation_min !=-1:\n",
    "            plt.imshow(activation[0,:,:, featuremap], interpolation=\"nearest\", vmin=activation_min, cmap=\"gray\")\n",
    "        else:\n",
    "            plt.imshow(activation[0,:,:, featuremap], interpolation=\"nearest\", cmap=\"gray\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "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": 1
}