arabic_handwriting (2).ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%reload_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai.vision import *\n",
    "from fastai.metrics import error_rate\n",
    "import csv\n",
    "import numpy as np\n",
    "import PIL \n",
    "import pandas as pd\n",
    "defaults.device = torch.device('cuda')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "#!pip install kaggle\n",
    "get_ipython().config.get('IPKernelApp', {})['parent_appname'] = \"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "00_notebook_tutorial.ipynb\t       lesson3-planet.ipynb\r\n",
      "ahdd1.zip\t\t\t       lesson4-collab.ipynb\r\n",
      "arabic_handwriting.ipynb\t       lesson4-tabular.ipynb\r\n",
      "Arabic Handwritten Digits Dataset CSV  lesson5-sgd-mnist.ipynb\r\n",
      "csvtest.csv\t\t\t       lesson6-pets-more.ipynb\r\n",
      "csvtestlabel.csv\t\t       lesson6-rossmann.ipynb\r\n",
      "csvtrain.csv\t\t\t       lesson7-human-numbers.ipynb\r\n",
      "csvtrainlabel.csv\t\t       lesson7-resnet-mnist.ipynb\r\n",
      "data\t\t\t\t       lesson7-superres-gan.ipynb\r\n",
      "digits\t\t\t\t       lesson7-superres-imagenet.ipynb\r\n",
      "images\t\t\t\t       lesson7-superres.ipynb\r\n",
      "lesson1-pets.ipynb\t\t       lesson7-wgan.ipynb\r\n",
      "lesson2-download.ipynb\t\t       my.png\r\n",
      "lesson2-sgd.ipynb\t\t       pix.png\r\n",
      "lesson3-camvid.ipynb\t\t       rossman_data_clean.ipynb\r\n",
      "lesson3-camvid-tiramisu.ipynb\t       test\r\n",
      "lesson3-head-pose.ipynb\t\t       Train + Test Matlab.mat\r\n",
      "lesson3-imdb.ipynb\r\n"
     ]
    }
   ],
   "source": [
    "!ls"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#ingest csv to pandas dataframe\n",
    "train = pd.read_csv('csvtrain.csv', nrows=2000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>0.1</th>\n",
       "      <th>0.2</th>\n",
       "      <th>0.3</th>\n",
       "      <th>0.4</th>\n",
       "      <th>0.5</th>\n",
       "      <th>0.6</th>\n",
       "      <th>0.7</th>\n",
       "      <th>0.8</th>\n",
       "      <th>0.9</th>\n",
       "      <th>...</th>\n",
       "      <th>0.486</th>\n",
       "      <th>0.487</th>\n",
       "      <th>0.488</th>\n",
       "      <th>0.489</th>\n",
       "      <th>0.490</th>\n",
       "      <th>0.491</th>\n",
       "      <th>0.492</th>\n",
       "      <th>0.493</th>\n",
       "      <th>0.494</th>\n",
       "      <th>0.495</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 784 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   0  0.1  0.2  0.3  0.4  0.5  0.6  0.7  0.8  0.9  ...  0.486  0.487  0.488  \\\n",
       "0  0    0    0    0    0    0    0    0    0    0  ...      0      0      0   \n",
       "1  0    0    0    0    0    0    0    0    0    0  ...      0      0      0   \n",
       "2  0    0    0    0    0    0    0    0    0    0  ...      0      0      0   \n",
       "3  0    0    0    0    0    0    0    0    0    0  ...      0      0      0   \n",
       "4  0    0    0    0    0    0    0    0    0    0  ...      0      0      0   \n",
       "\n",
       "   0.489  0.490  0.491  0.492  0.493  0.494  0.495  \n",
       "0      0      0      0      0      0      0      0  \n",
       "1      0      0      0      0      0      0      0  \n",
       "2      0      0      0      0      0      0      0  \n",
       "3      0      0      0      0      0      0      0  \n",
       "4      0      0      0      0      0      0      0  \n",
       "\n",
       "[5 rows x 784 columns]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "#uses pandas\n",
    "def pdMakePlot(row):\n",
    "    pixels = np.array(train.iloc[[row]], dtype='uint8')\n",
    "    pixels = pixels.reshape((28, 28)).T\n",
    "    pixels = np.true_divide(pixels, 255)\n",
    "    dim2 = np.zeros((28,28))\n",
    "    dim3 = np.zeros((28,28))\n",
    "    pix = np.stack((pixels, dim2,dim3), axis=2)\n",
    "    row += 1\n",
    "    filename = \"digits/%s.png\" % row\n",
    "    plt.imsave(filename, pix)\n",
    "#    plt.imshow(pix)\n",
    "    plt.close('all')\n",
    "    return"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>0.1</th>\n",
       "      <th>0.2</th>\n",
       "      <th>0.3</th>\n",
       "      <th>0.4</th>\n",
       "      <th>0.5</th>\n",
       "      <th>0.6</th>\n",
       "      <th>0.7</th>\n",
       "      <th>0.8</th>\n",
       "      <th>0.9</th>\n",
       "      <th>...</th>\n",
       "      <th>0.486</th>\n",
       "      <th>0.487</th>\n",
       "      <th>0.488</th>\n",
       "      <th>0.489</th>\n",
       "      <th>0.490</th>\n",
       "      <th>0.491</th>\n",
       "      <th>0.492</th>\n",
       "      <th>0.493</th>\n",
       "      <th>0.494</th>\n",
       "      <th>0.495</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1 rows × 784 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   0  0.1  0.2  0.3  0.4  0.5  0.6  0.7  0.8  0.9  ...  0.486  0.487  0.488  \\\n",
       "5  0    0    0    0    0    0    0    0    0    0  ...      0      0      0   \n",
       "\n",
       "   0.489  0.490  0.491  0.492  0.493  0.494  0.495  \n",
       "5      0      0      0      0      0      0      0  \n",
       "\n",
       "[1 rows x 784 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#load train data\n",
    "trainrows = 4000\n",
    "x = 5\n",
    "train = pd.read_csv('csvtrain.csv', nrows=trainrows)\n",
    "train.iloc[[x]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>0.1</th>\n",
       "      <th>0.2</th>\n",
       "      <th>0.3</th>\n",
       "      <th>0.4</th>\n",
       "      <th>0.5</th>\n",
       "      <th>0.6</th>\n",
       "      <th>0.7</th>\n",
       "      <th>0.8</th>\n",
       "      <th>0.9</th>\n",
       "      <th>...</th>\n",
       "      <th>0.538</th>\n",
       "      <th>0.539</th>\n",
       "      <th>0.540</th>\n",
       "      <th>0.541</th>\n",
       "      <th>0.542</th>\n",
       "      <th>0.543</th>\n",
       "      <th>0.544</th>\n",
       "      <th>0.545</th>\n",
       "      <th>0.546</th>\n",
       "      <th>0.547</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1 rows × 784 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   0  0.1  0.2  0.3  0.4  0.5  0.6  0.7  0.8  0.9  ...  0.538  0.539  0.540  \\\n",
       "3  0    0    0    0    0    0    0    0    0    0  ...      0      0      0   \n",
       "\n",
       "   0.541  0.542  0.543  0.544  0.545  0.546  0.547  \n",
       "3      0      0      0      0      0      0      0  \n",
       "\n",
       "[1 rows x 784 columns]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#load test data\n",
    "testrows = 5\n",
    "x = 3\n",
    "test = pd.read_csv('csvtest.csv', nrows=testrows) \n",
    "test.iloc[[x]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "csv = np.genfromtxt ('csvtrainlabel.csv', delimiter=\",\")\n",
    "csv = csv[0:trainrows]\n",
    "csv = csv.astype('int32')\n",
    "csv = np.add(csv,1)\n",
    "csv[csv == 10] = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1, 2, 3, 4, ..., 7, 8, 9, 0], dtype=int32)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['1', '2', '3', '4', ..., '3997', '3998', '3999', '4000'], dtype='<U21')"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#np array that we'll make into the filenames\n",
    "#from 1 to trainrows\n",
    "trainrange = trainrows +1\n",
    "files = np.arange(1,trainrange)\n",
    "files = files.astype(str)\n",
    "files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4000"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(files)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "#convert to filenames\n",
    "i = 0;\n",
    "j = 1;\n",
    "for file in files:\n",
    "    files[i] = \"%s.png\" % j\n",
    "    i += 1\n",
    "    j += 1\n",
    "    if i >= trainrange: break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['1.png', '2.png', '3.png', '4.png', ..., '3997.png', '3998.png', '3999.png', '4000.png'], dtype='<U21')"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#confirm the filenames in the array\n",
    "files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>name</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1.png</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2.png</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3.png</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4.png</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5.png</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    name  label\n",
       "0  1.png      1\n",
       "1  2.png      2\n",
       "2  3.png      3\n",
       "3  4.png      4\n",
       "4  5.png      5"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#combine two arrays into dataframe and add header\n",
    "df = pd.DataFrame({'name':files, 'label':csv})\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "i = 0\n",
    "max = trainrange-1\n",
    "for x in range(i,max):\n",
    "    pdMakePlot(i)\n",
    "    i += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "bs = 24"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'/home/ec2-user/SageMaker/course-v3/nbs/dl1/digits/'"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "path = '/home/ec2-user/SageMaker/course-v3/nbs/dl1/digits/'\n",
    "path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<bound method ImageDataBunch.from_df of <class 'fastai.vision.data.ImageDataBunch'>>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ImageDataBunch.from_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "tfms = get_transforms(do_flip=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = ImageDataBunch.from_df(path=path, df = df, ds_tfms=tfms, size=24)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "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\n",
      "image/png": "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\n",
      "text/plain": [
       "Image (3, 28, 28)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img = open_image('/home/ec2-user/SageMaker/course-v3/nbs/dl1/digits/3999.png')\n",
    "img"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 504x432 with 9 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "data.show_batch(rows=3, figsize=(7,6))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]\n"
     ]
    }
   ],
   "source": [
    "print(data.classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn = create_cnn(data, models.resnet34, metrics=accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Sequential(\n",
       "  (0): Sequential(\n",
       "    (0): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)\n",
       "    (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    (2): ReLU(inplace)\n",
       "    (3): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
       "    (4): Sequential(\n",
       "      (0): BasicBlock(\n",
       "        (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "      (1): BasicBlock(\n",
       "        (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "      (2): BasicBlock(\n",
       "        (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "    )\n",
       "    (5): Sequential(\n",
       "      (0): BasicBlock(\n",
       "        (conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (downsample): Sequential(\n",
       "          (0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
       "          (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        )\n",
       "      )\n",
       "      (1): BasicBlock(\n",
       "        (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "      (2): BasicBlock(\n",
       "        (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "      (3): BasicBlock(\n",
       "        (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "    )\n",
       "    (6): Sequential(\n",
       "      (0): BasicBlock(\n",
       "        (conv1): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (downsample): Sequential(\n",
       "          (0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
       "          (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        )\n",
       "      )\n",
       "      (1): BasicBlock(\n",
       "        (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "      (2): BasicBlock(\n",
       "        (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "      (3): BasicBlock(\n",
       "        (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "      (4): BasicBlock(\n",
       "        (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "      (5): BasicBlock(\n",
       "        (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "    )\n",
       "    (7): Sequential(\n",
       "      (0): BasicBlock(\n",
       "        (conv1): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (downsample): Sequential(\n",
       "          (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
       "          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        )\n",
       "      )\n",
       "      (1): BasicBlock(\n",
       "        (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "      (2): BasicBlock(\n",
       "        (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace)\n",
       "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "    )\n",
       "  )\n",
       "  (1): Sequential(\n",
       "    (0): AdaptiveConcatPool2d(\n",
       "      (ap): AdaptiveAvgPool2d(output_size=1)\n",
       "      (mp): AdaptiveMaxPool2d(output_size=1)\n",
       "    )\n",
       "    (1): Flatten()\n",
       "    (2): BatchNorm1d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    (3): Dropout(p=0.25)\n",
       "    (4): Linear(in_features=1024, out_features=512, bias=True)\n",
       "    (5): ReLU(inplace)\n",
       "    (6): BatchNorm1d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    (7): Dropout(p=0.5)\n",
       "    (8): Linear(in_features=512, out_features=10, bias=True)\n",
       "  )\n",
       ")"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learn.model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:16 <p><table style='width:300px; margin-bottom:10px'>\n",
       "  <tr>\n",
       "    <th>epoch</th>\n",
       "    <th>train_loss</th>\n",
       "    <th>valid_loss</th>\n",
       "    <th>accuracy</th>\n",
       "  </tr>\n",
       "  <tr>\n",
       "    <th>1</th>\n",
       "    <th>1.461245</th>\n",
       "    <th>0.665243</th>\n",
       "    <th>0.802500</th>\n",
       "  </tr>\n",
       "  <tr>\n",
       "    <th>2</th>\n",
       "    <th>0.790824</th>\n",
       "    <th>0.228773</th>\n",
       "    <th>0.931250</th>\n",
       "  </tr>\n",
       "  <tr>\n",
       "    <th>3</th>\n",
       "    <th>0.502614</th>\n",
       "    <th>0.166105</th>\n",
       "    <th>0.961250</th>\n",
       "  </tr>\n",
       "  <tr>\n",
       "    <th>4</th>\n",
       "    <th>0.387829</th>\n",
       "    <th>0.167723</th>\n",
       "    <th>0.955000</th>\n",
       "  </tr>\n",
       "</table>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
     ]
    }
   ],
   "source": [
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:15 <p><table style='width:300px; margin-bottom:10px'>\n",
       "  <tr>\n",
       "    <th>epoch</th>\n",
       "    <th>train_loss</th>\n",
       "    <th>valid_loss</th>\n",
       "    <th>accuracy</th>\n",
       "  </tr>\n",
       "  <tr>\n",
       "    <th>1</th>\n",
       "    <th>0.287015</th>\n",
       "    <th>0.307796</th>\n",
       "    <th>0.928750</th>\n",
       "  </tr>\n",
       "  <tr>\n",
       "    <th>2</th>\n",
       "    <th>0.221842</th>\n",
       "    <th>0.030812</th>\n",
       "    <th>0.991250</th>\n",
       "  </tr>\n",
       "  <tr>\n",
       "    <th>3</th>\n",
       "    <th>0.124907</th>\n",
       "    <th>0.009651</th>\n",
       "    <th>0.996250</th>\n",
       "  </tr>\n",
       "</table>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.unfreeze()\n",
    "learn.fit_one_cycle(3, max_lr=slice(.006, .004))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0, 0, 0, 0, ..., 0, 0, 0, 0]])"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test = pd.read_csv('csvtest.csv', nrows=5)\n",
    "test2 = test.iloc[[3]]\n",
    "test2.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "pixels = np.array(test2.values, dtype='uint8')\n",
    "pixels = pixels.reshape((28, 28)).T\n",
    "pixels = np.true_divide(pixels, 255)\n",
    "dim2 = np.zeros((28,28))\n",
    "dim3 = np.zeros((28,28))\n",
    "pix = np.stack((pixels, dim2,dim3), axis=2)\n",
    "filename = \"test/%s.png\" % 3\n",
    "plt.imsave(filename, pix)\n",
    "#    plt.imshow(pix)\n",
    "plt.close('all')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "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\n",
      "image/png": "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\n",
      "text/plain": [
       "Image (3, 28, 28)"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img = open_image('/home/ec2-user/SageMaker/course-v3/nbs/dl1/test/3.png')\n",
    "img"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "#now to production\n",
    "defaults.device = torch.device('cpu')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Category 4"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#learn = load_learner(path)\n",
    "pred_class,pred_idx,outputs = learn.predict(img)\n",
    "pred_class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "fastai"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}