Traffic_Sign_Classifier-final.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"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "## Step 0: Load The Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "finished\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",
    "\n",
    "training_file = 'train.p'\n",
    "validation_file='valid.p'\n",
    "testing_file = '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_valid, y_valid = valid['features'], valid['labels']\n",
    "X_test, y_test = test['features'], test['labels']\n",
    "\n",
    "print('finished')"
   ]
  },
  {
   "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": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of training examples = 34799\n",
      "Number of testing examples = 12630\n",
      "Number of validation examples  4410\n",
      "Image data shape = (32, 32, 3)\n",
      "Number of classes = 43\n"
     ]
    }
   ],
   "source": [
    "\n",
    "# TODO: Number of training examples\n",
    "n_train = X_train.shape[0]\n",
    "\n",
    "# TODO: Number of validation examples\n",
    "n_validation = X_valid.shape[0]\n",
    "\n",
    "# TODO: Number of testing examples.\n",
    "n_test = X_test.shape[0]\n",
    "\n",
    "# TODO: What's the shape of an traffic sign image?\n",
    "image_shape = X_train[0].shape\n",
    "\n",
    "# TODO: How many unique classes/labels there are in the dataset.\n",
    "n_classes = len(set(y_train))\n",
    "\n",
    "print(\"Number of training examples =\", n_train)\n",
    "print(\"Number of testing examples =\", n_test)\n",
    "print(\"Number of validation examples \",n_validation)\n",
    "print(\"Image data shape =\", image_shape)\n",
    "print(\"Number of classes =\", n_classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Include an exploratory visualization of the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Image Dimensions:  (32, 32, 3)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x20d303b94e0>"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAB2CAYAAADY3GjsAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x20d2d76a5f8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import cv2\n",
    "import numpy as np\n",
    "%matplotlib inline\n",
    "\n",
    "\n",
    "print('Image Dimensions: ',X_train[1500].shape)\n",
    "plt.imshow(np.hstack((X_train[14800], X_train[1500], X_train[2500], X_train[3500] ) ))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "----\n",
    "\n",
    "## Step 2: Design and Test a Model Architecture\n",
    "\n"
   ]
  },
  {
   "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": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "finished\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "for i in range(X_train.shape[0]):\n",
    "    image = np.zeros((32,32,3) )\n",
    "    image[:,:,0]=cv2.equalizeHist(X_train[i,:,:,0])\n",
    "    image[:,:,1]=cv2.equalizeHist(X_train[i,:,:,1])\n",
    "    image[:,:,2]=cv2.equalizeHist(X_train[i,:,:,2])\n",
    "    X_train[i,:,:,:]=image\n",
    "\n",
    "for i in range(X_valid.shape[0]):\n",
    "    image = np.zeros((32,32,3))\n",
    "    image[:,:,0]=cv2.equalizeHist(X_valid[i,:,:,0])\n",
    "    image[:,:,1]=cv2.equalizeHist(X_valid[i,:,:,1])\n",
    "    image[:,:,2]=cv2.equalizeHist(X_valid[i,:,:,2])\n",
    "    X_valid[i,:,:,:]=image\n",
    "    \n",
    "for i in range(X_test.shape[0]):\n",
    "    image = np.zeros((32,32,3))\n",
    "    image[:,:,0]=cv2.equalizeHist(X_test[i,:,:,0])\n",
    "    image[:,:,1]=cv2.equalizeHist(X_test[i,:,:,1])\n",
    "    image[:,:,2]=cv2.equalizeHist(X_test[i,:,:,2])\n",
    "    X_test[i,:,:,:]=image\n",
    "    \n",
    "    \n",
    "X_train = np.asfarray(X_train, dtype='float32')\n",
    "X_valid = np.asfarray(X_valid, dtype='float32')\n",
    "X_test = np.asfarray(X_test, dtype='float32')\n",
    "\n",
    "for i in range(X_train.shape[0]):\n",
    "    X_train[i,:,:,:]=(X_train[i,:,:,:]-128)/128\n",
    "for i in range(X_valid.shape[0]):\n",
    "    X_valid[i,:,:,:]=(X_valid[i,:,:,:]-128)/128\n",
    "for i in range(X_test.shape[0]):\n",
    "    X_test[i,:,:,:]=(X_test[i,:,:,:]-128)/128\n",
    "\n",
    "\n",
    "print('finished')\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model Architecture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow.contrib.layers import flatten\n",
    "from sklearn.utils import shuffle\n",
    "EPOCHS = 12\n",
    "BATCH_SIZE = 128\n",
    "\n",
    "def LeNet(x):    \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",
    "    W1= tf.Variable(tf.truncated_normal([5,5,3,9],mean=mu,stddev=sigma,dtype=tf.float32,seed=None,name=None))\n",
    "    b1= tf.Variable(tf.zeros(9))\n",
    "    \n",
    "    W2= tf.Variable(tf.truncated_normal([5,5,9,16],mean=mu,stddev=sigma,dtype=tf.float32,seed=None,name=None))\n",
    "    b2= tf.Variable(tf.zeros(16))\n",
    "    \n",
    "    w3= tf.Variable(tf.truncated_normal([400,120],mean=mu,stddev=sigma,dtype=tf.float32,seed=None,name=None))\n",
    "    b3= tf.Variable(tf.zeros(120))\n",
    "    \n",
    "    w4= tf.Variable(tf.truncated_normal([120,84],mean=mu,stddev=sigma,dtype=tf.float32,seed=None,name=None))\n",
    "    b4= tf.Variable(tf.zeros(84))\n",
    "    \n",
    "    w5= tf.Variable(tf.truncated_normal([84,43],mean=mu,stddev=sigma,dtype=tf.float32,seed=None,name=None))\n",
    "    b5= tf.Variable(tf.zeros(43))\n",
    "   \n",
    "    # TODO: Layer 1: Convolutional. Input = 32x32x3. Output = 25x25x9.\n",
    "    conv1 = tf.nn.conv2d(x,W1,strides=[1,1,1,1],padding='VALID')\n",
    "    conv1 = tf.nn.bias_add(conv1,b1)\n",
    "    # TODO: Activation.\n",
    "    conv1=tf.nn.relu(conv1)\n",
    "    # TODO: Pooling. Input = 25x25x9. Output = 11x11x9.\n",
    "    p1 = tf.nn.max_pool(conv1,ksize=[1,2,2,1],strides=[1,2,2,1],padding='VALID')\n",
    "    # TODO: Layer 2: Convolutional. Output = 7x7x16.\n",
    "    conv2 = tf.nn.conv2d(p1,W2,strides=[1,1,1,1],padding='VALID')\n",
    "    conv2 = tf.nn.bias_add(conv2,b2)\n",
    "    # TODO: Activation.\n",
    "    conv2=tf.nn.relu(conv2)\n",
    "    # TODO: Pooling. Input = 7x7x16. Output = 4x4x16.\n",
    "    p2 = tf.nn.max_pool(conv2,ksize=[1,2,2,1],strides=[1,2,2,1],padding='VALID')\n",
    "    # TODO: Flatten. Input = 5x5x16. Output =256.\n",
    "    flat = flatten(p2)\n",
    "    # TODO: Layer 3: Fully Connected. Input = 192. Output = 120.\n",
    "    fc1 = tf.add(tf.matmul(flat,w3),b3)\n",
    "    # TODO: Activation.\n",
    "    fc1 = tf.nn.relu(fc1)\n",
    "    #fc1 = tf.nn.dropout(fc1, keep_prob=0.5)\n",
    "    # TODO: Layer 4: Fully Connected. Input = 120. Output = 84.\n",
    "    fc2 = tf.add(tf.matmul(fc1,w4),b4)\n",
    "    # TODO: Activation.\n",
    "    fc2 = tf.nn.relu(fc2)\n",
    "    # TODO: Layer 5: Fully Connected. Input = 84. Output = 43.\n",
    "    logits = tf.add(tf.matmul(fc2,w5),b5)\n",
    "    return logits\n",
    "\n",
    "x = tf.placeholder(tf.float32,(None,32,32,3))\n",
    "y = tf.placeholder(tf.int32, (None))\n",
    "one_hot_y = tf.one_hot(y, 43)\n",
    "\n",
    "logits =LeNet(x)\n",
    "cross_entropy = tf.nn.softmax_cross_entropy_with_logits(labels=one_hot_y,logits=logits)\n",
    "loss = tf.reduce_mean(cross_entropy)\n",
    "optimizer = tf.train.AdamOptimizer(learning_rate=0.001)\n",
    "training_operation = optimizer.minimize(loss)\n",
    "\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})\n",
    "        total_accuracy += (accuracy * len(batch_x))\n",
    "    return total_accuracy / num_examples"
   ]
  },
  {
   "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": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training...\n",
      "\n",
      "EPOCH 1 ...\n",
      "Validation Accuracy = 0.815\n",
      "\n",
      "EPOCH 2 ...\n",
      "Validation Accuracy = 0.889\n",
      "\n",
      "EPOCH 3 ...\n",
      "Validation Accuracy = 0.919\n",
      "\n",
      "EPOCH 4 ...\n",
      "Validation Accuracy = 0.921\n",
      "\n",
      "EPOCH 5 ...\n",
      "Validation Accuracy = 0.912\n",
      "\n",
      "EPOCH 6 ...\n",
      "Validation Accuracy = 0.930\n",
      "\n",
      "EPOCH 7 ...\n",
      "Validation Accuracy = 0.934\n",
      "\n",
      "EPOCH 8 ...\n",
      "Validation Accuracy = 0.934\n",
      "\n",
      "EPOCH 9 ...\n",
      "Validation Accuracy = 0.940\n",
      "\n",
      "EPOCH 10 ...\n",
      "Validation Accuracy = 0.943\n",
      "\n",
      "EPOCH 11 ...\n",
      "Validation Accuracy = 0.937\n",
      "\n",
      "EPOCH 12 ...\n",
      "Validation Accuracy = 0.936\n",
      "\n",
      "Model saved\n",
      "Test Accuracy = 0.925\n"
     ]
    }
   ],
   "source": [
    "\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})\n",
    "            \n",
    "        validation_accuracy = evaluate(X_valid, y_valid)\n",
    "        print(\"EPOCH {} ...\".format(i+1))\n",
    "        print(\"Validation Accuracy = {:.3f}\".format(validation_accuracy))\n",
    "        print()\n",
    "        \n",
    "    saver.save(sess, './mysession')\n",
    "    print(\"Model saved\")\n",
    "    test_accuracy = evaluate(X_test, y_test)\n",
    "    print(\"Test Accuracy = {:.3f}\".format(test_accuracy))\n",
    "\n",
    "\n"
   ]
  },
  {
   "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": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x1d0e0ab6c18>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAABlCAYAAABdnhjZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nBQ/0wAILLLBdxIIHemCBBRbYLmL/H/GOExpjpbF9AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1d0e0766908>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "import matplotlib.image as mpimg\n",
    "img1 = np.asfarray(mpimg.imread('new_test_images/1.jpg'), dtype='float32')\n",
    "img3 = np.asfarray(mpimg.imread('new_test_images/3.jpg'), dtype='float32')\n",
    "img11 = np.asfarray(mpimg.imread('new_test_images/11.jpg'), dtype='float32')\n",
    "img35 = np.asfarray(mpimg.imread('new_test_images/35.jpg'), dtype='float32')\n",
    "img38 = np.asfarray(mpimg.imread('new_test_images/38.jpg'), dtype='float32')\n",
    "\n",
    "\n",
    "\n",
    "all_images = np.hstack((img1,img3,img11,img35,img38))\n",
    "\n",
    "img1=(img1-128)/128\n",
    "img3=(img3-128)/128\n",
    "img11=(img11-128)/128\n",
    "img35=(img35-128)/128\n",
    "img38=(img38-128)/128\n",
    "\n",
    "new_X_test = [img1,img3,img11,img35,img38]\n",
    "new_y_test=[1,3,11,35,38]\n",
    "\n",
    "\n",
    "\n",
    "plt.imshow(all_images)\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predict the Sign Type for Each Image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from .\\mysession\n",
      "Test Accuracy = 0.925\n",
      "prediction result of first image:  [ True]\n",
      "prediction result of second image:  [ True]\n",
      "prediction result of third image:  [ True]\n",
      "prediction result of fourth image:  [ True]\n",
      "prediction result of fifth image:  [ True]\n"
     ]
    }
   ],
   "source": [
    "### 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",
    "with tf.Session() as sess:\n",
    "    new_saver = tf.train.import_meta_graph(\"mysession.meta\")\n",
    "    new_saver.restore(sess, tf.train.latest_checkpoint('.'))\n",
    "\n",
    "    accuracy = evaluate(X_test,y_test)\n",
    "    print(\"Test Accuracy = {:.3f}\".format(accuracy))\n",
    "    prediction = tf.equal(tf.argmax(logits, 1), tf.argmax(one_hot_y, 1))\n",
    "    result = sess.run(prediction,feed_dict={x:[img1],y:[1]})\n",
    "    print('prediction result of first image: ',result)\n",
    "    result = sess.run(prediction,feed_dict={x:[img3],y:[3]})\n",
    "    print('prediction result of second image: ',result)\n",
    "    result = sess.run(prediction,feed_dict={x:[img11],y:[11]})\n",
    "    print('prediction result of third image: ',result)\n",
    "    result = sess.run(prediction,feed_dict={x:[img35],y:[35]})\n",
    "    print('prediction result of fourth image: ',result)\n",
    "    result = sess.run(prediction,feed_dict={x:[img38],y:[38]})\n",
    "    print('prediction result of fifth image: ',result)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Analyze Performance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from .\\mysession\n",
      "New images test accuracy = 100.000%\n"
     ]
    }
   ],
   "source": [
    "### Calculate the accuracy for these 5 new images. \n",
    "### For example, if the model predicted 1 out of 5 signs correctly, it's 20% accurate on these new images.\n",
    "with tf.Session() as sess:\n",
    "    new_saver = tf.train.import_meta_graph(\"mysession.meta\")\n",
    "    new_saver.restore(sess, tf.train.latest_checkpoint('.'))\n",
    "    accuracy = evaluate(new_X_test,new_y_test)\n",
    "    print(\"New images test accuracy = {:.3f}%\".format(accuracy*100))\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": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from .\\mysession\n",
      "TopKV2(values=array([[  9.99933839e-01,   6.23782325e-05,   3.27759949e-06,\n",
      "          4.64077118e-07,   1.56517544e-07],\n",
      "       [  9.82837081e-01,   1.71621814e-02,   7.34060279e-07,\n",
      "          1.79706330e-08,   1.47235624e-09],\n",
      "       [  9.85541403e-01,   1.18939187e-02,   2.46886560e-03,\n",
      "          9.54436255e-05,   5.25370865e-07],\n",
      "       [  9.99999642e-01,   2.60439634e-07,   8.69375327e-08,\n",
      "          1.92123140e-09,   2.29829680e-11],\n",
      "       [  9.99999762e-01,   2.04484422e-07,   8.54978657e-11,\n",
      "          5.45021840e-11,   1.12779881e-11]], dtype=float32), indices=array([[ 1,  0,  5,  4,  2],\n",
      "       [ 3,  2,  5, 31, 19],\n",
      "       [11, 30, 27, 18, 20],\n",
      "       [35, 40, 36, 33, 34],\n",
      "       [38, 40, 36, 34, 12]]))\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",
    "    new_saver = tf.train.import_meta_graph(\"mysession.meta\")\n",
    "    new_saver.restore(sess, tf.train.latest_checkpoint('.'))\n",
    "    softmax_propabilities = tf.nn.softmax(logits,-1)\n",
    "    top_five = tf.nn.top_k(softmax_propabilities,k=5)\n",
    "    output = sess.run(top_five,feed_dict={x:new_X_test})\n",
    "    print(output)\n",
    "    \n",
    "    \n"
   ]
  }
 ],
 "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
}