Image_Recognizer_02.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Using our Json model to predict an image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [],
   "source": [
    "# importing the libraries\n",
    "import keras\n",
    "from keras.models import model_from_json\n",
    "from pathlib import Path\n",
    "from keras.preprocessing import image\n",
    "import numpy as np\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [],
   "source": [
    "# intialize the class labels for the cifar10 dataset\n",
    "class_labels = [\n",
    "    \"Plane\",\n",
    "    \"Cat\",\n",
    "    \"Bird\",\n",
    "    \"Deer\",\n",
    "    \"Dog\",\n",
    "    \"Frog\",\n",
    "    \"Horse\",\n",
    "    \"Boat\",\n",
    "    \"Truck\"\n",
    "    ]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load the json file that contain neural structure\n",
    "f = Path(\"model_structure.json\")\n",
    "model_structure = f.read_text()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [],
   "source": [
    "# recreate the model from the json file\n",
    "model = model_from_json(model_structure)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [],
   "source": [
    "#reload the weights to the neurons\n",
    "model.load_weights(\"model_weights.h5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load the image file\n",
    "img = image.load_img(\"bird.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [],
   "source": [
    "# convert to the image array\n",
    "img_to_test = image.img_to_array(img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Add the fourth dimenstion\n",
    "list_of_images = np.expand_dims(img_to_test, axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [],
   "source": [
    "# make prediction using model\n",
    "results = model.predict(list_of_images)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [],
   "source": [
    "# First result\n",
    "single_result = results[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [],
   "source": [
    "# find likelyhood values for the first 10 possible classes\n",
    "most_likely_class_index = int(np.argmax(single_result))\n",
    "class_likelyhood = single_result[most_likely_class_index]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [],
   "source": [
    "# get the names of the most likely class\n",
    "class_label = class_labels[most_likely_class_index]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "This image is a Bird - Likelyhood: 1.000000\n"
     ]
    }
   ],
   "source": [
    "# Print the result\n",
    "print(\"This image is a {} - Likelyhood: {:2f}\".format(class_label, class_likelyhood))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}