CNN_with_fashion_mnist.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "CNN with fashion_mnist.ipynb",
      "provenance": [],
      "authorship_tag": "ABX9TyMKfeI7SsSvw/U0auKbtukG",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/GordanY/tensorflow-learning/blob/master/CNN_with_fashion_mnist.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "L8qjWAPpGxXh",
        "colab_type": "text"
      },
      "source": [
        "## Another example of CNN using fashion mnist datatsets\n",
        "\n",
        "Identical coding with slightly modified mormalization function"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JLsJUtY_DHay",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import tensorflow as tf\n",
        "import tensorflow_datasets as tfds\n",
        "import math\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "k9PRaaUdHCIl",
        "colab_type": "text"
      },
      "source": [
        "## Get the sample"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Z4RcAWkiDXP4",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "ds, md = tfds.load(\"fashion_mnist\", as_supervised = True, with_info = True)\n",
        "train_ds, test_ds = ds[\"train\"], ds[\"test\"]"
      ],
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6DiKPT1zE5tv",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def normalize(datas, labels):\n",
        "  datas = tf.cast(datas, tf.float32) / 255\n",
        "  return datas, labels\n",
        "\n",
        "\n",
        "train_ds, test_ds = train_ds.map(normalize), test_ds.map(normalize)"
      ],
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zHogRlCuHIaf",
        "colab_type": "text"
      },
      "source": [
        "## Train the model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mByt6VzxE-5S",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model = tf.keras.Sequential([\n",
        "    tf.keras.layers.Conv2D(32, (3,3), padding='same', activation=tf.nn.relu,\n",
        "                           input_shape=(28, 28, 1)),\n",
        "    tf.keras.layers.MaxPooling2D((2, 2), strides=2),\n",
        "    tf.keras.layers.Conv2D(64, (3,3), padding='same', activation=tf.nn.relu),\n",
        "    tf.keras.layers.MaxPooling2D((2, 2), strides=2),\n",
        "    tf.keras.layers.Flatten(),\n",
        "    tf.keras.layers.Dense(128, activation=tf.nn.relu),\n",
        "    tf.keras.layers.Dense(10, activation=tf.nn.softmax)\n",
        "])"
      ],
      "execution_count": 17,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RQNErNSuFCMX",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.compile(optimizer='adam',\n",
        "              loss=tf.keras.losses.SparseCategoricalCrossentropy(),\n",
        "              metrics=['accuracy'])"
      ],
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZWu924L1FGkk",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "BATCH_SIZE = 32\n",
        "train_ds = train_ds.cache().repeat().shuffle(md.splits[\"train\"].num_examples).batch(BATCH_SIZE)\n",
        "test_ds = test_ds.cache().batch(BATCH_SIZE)"
      ],
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EqQIZt9-FKSc",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 207
        },
        "outputId": "1aa10c69-7594-4803-b942-089df9915c41"
      },
      "source": [
        "history = model.fit(train_ds, epochs=5, steps_per_epoch=math.ceil(md.splits[\"train\"].num_examples/BATCH_SIZE))"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/5\n",
            "1875/1875 [==============================] - 11s 6ms/step - loss: 0.3973 - accuracy: 0.8564\n",
            "Epoch 2/5\n",
            "1875/1875 [==============================] - 10s 5ms/step - loss: 0.2645 - accuracy: 0.9040\n",
            "Epoch 3/5\n",
            "1875/1875 [==============================] - 10s 5ms/step - loss: 0.2203 - accuracy: 0.9194\n",
            "Epoch 4/5\n",
            "1875/1875 [==============================] - 10s 5ms/step - loss: 0.1862 - accuracy: 0.9313\n",
            "Epoch 5/5\n",
            "1875/1875 [==============================] - 10s 5ms/step - loss: 0.1629 - accuracy: 0.9399\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "68e681jaF-sh",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 296
        },
        "outputId": "40f88739-1e0f-49ca-85d7-7de996c33d9c"
      },
      "source": [
        "plt.xlabel('Epoch Number')\n",
        "plt.ylabel(\"Loss Magnitude\")\n",
        "plt.plot(history.history['loss'])"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<matplotlib.lines.Line2D at 0x7f05263e1a58>]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 21
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEGCAYAAAB/+QKOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3deXhU9d3+8fcnCftOCGsChE0EWZO6QEXA3VrcwGq1Vbv4WNc+2rq1tlZr61KXWrXW9mcf21qtIFa07iDuCwmyLxL2sO9hCyHJ5/fHHDSmkzCBzJxJcr+uay5nzpwzc1+nndyc7XvM3REREaksJewAIiKSnFQQIiISlQpCRESiUkGIiEhUKggREYkqLewAtaVDhw7es2fPsGOIiNQp+fn5m909I9p79aYgevbsSV5eXtgxRETqFDNbWdV72sUkIiJRqSBERCQqFYSIiEQV14Iws9PMbLGZFZjZzdXMd56ZuZnlVph2S7DcYjM7NZ45RUTkv8XtILWZpQKPAicDhcAMM5vi7gsqzdcKuA74pMK0AcAFwECgK/CWmfVz97J45RURka+K5xbE0UCBuy9z9xLgWeCsKPPdCdwDFFeYdhbwrLvvc/flQEHweSIikiDxLIhuwOoKrwuDaV8ws+FAlrv/p6bLBstfbmZ5Zpa3adOm2kktIiJAiAepzSwFeAC44VA/w92fcPdcd8/NyIh6ncdBFe8v47evLmT11j2HGkNEpF6KZ0GsAbIqvM4Mph3QCjgKmG5mK4BjgSnBgeqDLVtrNu/ax9Mfr+LGSXMoL9e9MUREDohnQcwA+ppZtpk1JnLQecqBN919h7t3cPee7t4T+BgY5+55wXwXmFkTM8sG+gKfxiNkZrvm3HbmkXy0bAt/+2hFPL5CRKROiltBuHspcDXwOrAQeM7d55vZHWY27iDLzgeeAxYArwFXxfMMpvNzsxhzRAZ3v7aIZZt2xetrRETqFKsvtxzNzc31wxmLaUNRMac8+C69Mlow6YoRpKZYLaYTEUlOZpbv7rnR3tOV1IFOrZtyx1kD+WzVdp54d1nYcUREQqeCqGDckK6cflRnHnzzcxav3xl2HBGRUKkgKjAzfn32UbRqmsb1z81if1l52JFEREKjgqgkvWUTfnPuIOavLeKRaQVhxxERCY0KIopTB3bm3GHdeOTtAuYW7gg7johIKFQQVfjlNweS0bIJ1z83i+L9GiNQRBoeFUQV2jRvxD3jB7Nk4y4efPPzsOOIiCScCqIaJ/TL4NvHdOeJ95aRt2Jr2HFERBJKBXEQt55xJJntmnHDxNnsKSkNO46ISMKoIA6iZZM07hs/hFVb93D3q4vCjiMikjAqiBgc2yud743M5m8freT9JZvDjiMikhAqiBj99NQj6JXRghsnzaaoeH/YcURE4k4FEaOmjVK5f8IQ1hcV8+uXFxx8ARGROk4FUQPDurfjR6N781xeIVMXbgg7johIXKkgaujaE/vSv3Mrbp48l227S8KOIyISNyqIGmqSlsoD5w9l+54SfjFlfthxRETiRgVxCAZ0bc11J/blpdlreXnO2rDjiIjEhQriEF1xQm+GZLbhtn/PY+PO4rDjiIjUOhXEIUpLTeH+84eyp6SMWyfPo77culVE5AAVxGHo07ElPz31CN5auIHnZ64JO46ISK1SQRym743M5ujs9vxqynzWbt8bdhwRkVqjgjhMKSnG78YPocydm56fo11NIlJvqCBqQff05tx6xpG8t2QzT3+yKuw4IiK1QgVRSy46pjvH9+3Ab15ZyMotu8OOIyJy2FQQtcTMuOe8waSmGD+dOIeycu1qEpG6TQVRi7q2bcbt3xzIpyu28tcPlocdR0TksKggatm5w7tx8oBO3Pv6Ygo27gw7jojIIVNB1DIz4zfnDKJF41Suf242pWXlYUcSETkkKog4yGjVhLvOGcScwh38cfrSsOOIiBwSFUScnDGoC+OGdOX3U5cwf+2OsOOIiNSYCiKO7jhrIO1aNOaG52azr7Qs7DgiIjWigoijts0bc895g1i0ficPT10SdhwRkRpRQcTZ2P6dOD83kz9OX8pnq7aFHUdEJGZxLQgzO83MFptZgZndHOX9K8xsrpnNMrP3zWxAML2nme0Nps8ys8fjmTPebjtzAF3aNOOG52azt0S7mkSkbohbQZhZKvAocDowALjwQAFU8E93H+TuQ4F7gQcqvLfU3YcGjyvilTMRWjVtxL3jB7Ns827ue31x2HFERGISzy2Io4ECd1/m7iXAs8BZFWdw96IKL1sA9XZ8ipF9OnDJcT148oPlfLR0S9hxREQOKp4F0Q1YXeF1YTDtK8zsKjNbSmQL4toKb2Wb2Wdm9o6ZHR/tC8zscjPLM7O8TZs21Wb2uLjp9P70TG/OTyfNZte+0rDjiIhUK/SD1O7+qLv3Bm4Cfh5MXgd0d/dhwPXAP82sdZRln3D3XHfPzcjISFzoQ9S8cRr3nz+Etdv3ctd/FoYdR0SkWvEsiDVAVoXXmcG0qjwLnA3g7vvcfUvwPB9YCvSLU86EyunRnh+O6sUzn65i+uKNYccREalSPAtiBtDXzLLNrDFwATCl4gxm1rfCy28AS4LpGcFBbsysF9AXWBbHrAn1vyf1o1+nltz0/Bx27NkfdhwRkajiVhDuXgpcDbwOLASec/f5ZnaHmY0LZrvazOab2Swiu5IuCaaPAuYE0ycBV7j71nhlTbSmjVK5f8JQtuwq4faX5ocdR0QkKqsv91DOzc31vLy8sGPUyENvfc5Dby3h8YtzOO2ozmHHEZEGyMzy3T032nuhH6RuyK4a04ejurXmZy/MZcuufWHHERH5ChVEiBqlpvDA+UPZWVzKz16YR33ZmhOR+kEFEbJ+nVpx/Sn9eG3+eqbMXht2HBGRL6ggksAPj+/F8O5tue3f89hQVBx2HBERQAWRFFJTjPvPH0pJWTk3PT9Hu5pEJCmoIJJEdocW3HL6kUxfvIl/zVh98AVEROJMBZFEvnNsD0b0TufOlxeweuuesOOISAOngkgiKSnGveMHY2b8dNJsysu1q0lEwqOCSDKZ7ZrzizMH8PGyrTz10Yqw44hIA6aCSEITcjMZ278j97y2iGWbdoUdR0QaKBVEEjIz7j53EE3SUrlh4mzKtKtJREKggkhSHVs35Y6zBvLZqu088W69GchWROoQFUQSGzekK2cM6syDb37OovVFB19ARKQWqSCSmJlx51lH0bpZGjc8N5uS0vKwI4lIA6KCSHLpLZtw1zmDmL+2iEfeLgg7jog0ICqIOuDUgZ05d3g3Hn27gDmF28OOIyINxEELwiIuNrNfBK+7m9nR8Y8mFf3ymwPJaNmE65+bTfH+srDjiEgDEMsWxGPAccCFweudwKNxSyRRtWnWiHvHD6Zg4y4eePPzsOOISAMQS0Ec4+5XAcUA7r4NaBzXVBLVqH4ZXHRMd/783jLyVtSbW3SLSJKKpSD2m1kq4ABmlgHodJqQ3HrGkWS2a8YNE2ezp6Q07DgiUo/FUhAPAy8AHc3sLuB94DdxTSVVatEkjd+NH8KqrXu4+9VFYccRkXos7WAzuPvTZpYPnAgYcLa7L4x7MqnSMb3S+f7IbP7y/nJOGdCZr/ftEHYkEamHqtyCMLP2Bx7ARuAZ4J/AhmCahOgnpx5B74wW3DhpNkXF+8OOIyL1UHW7mPKBvOC/m4DPgSXB8/z4R5PqNG2Uyv3nD2V9UTF3vrQg7DgiUg9VWRDunu3uvYC3gG+6ewd3TwfOBN5IVECp2tCstlw5ug8T8wt5a8GGsOOISD0Ty0HqY939lQMv3P1VYET8IklNXHtiX/p3bsXNk+eybXdJ2HFEpB6JpSDWmtnPzaxn8PgZsDbewSQ2jdNSeOD8oezYW8JtL84LO46I1COxFMSFQAaRU11fADry5VXVkgQGdG3Nj0/qx8tz1vHyHHW3iNSOWE5z3Qpcl4Aschj+Z1Qv3liwgdv+PY+js9vTsVXTsCOJSB0Xy2B9b5vZtMqPRIST2KWlpnD/hCHsKSnj1snzcNdtSkXk8Bx0CwL4SYXnTYHzAI3xkIT6dGzJjaf1586XF/D8zDWMz8kMO5KI1GGx7GKqfM3DB2b2aZzyyGG6bERP3pi/nl9Nmc+I3ul0bdss7EgiUkfFsoupfYVHBzM7FWiTgGxyCFJSjPvGD6HMnRsnzdGuJhE5ZLGcxVTxiuqPgBuA78fy4WZ2mpktNrMCM7s5yvtXmNlcM5tlZu+b2YAK790SLLc4KCWJUff05vzsG0fyfsFm/vHJqrDjiEgdFcsxiCPdvbjiBDNrcrCFgiHCHwVOBgqBGWY2xd0rjgvxT3d/PJh/HPAAcFpQFBcAA4GuwFtm1s/ddSu1GH376O68Nm89v/nPQkb17UCP9BZhRxKROiaWLYgPo0z7KIbljgYK3H2Zu5cAzwJnVZzB3YsqvGxBcM+JYL5n3X2fuy8HCoLPkxiZGfeOH0xaqvGTibMpK9euJhGpmepGc+1sZjlAMzMbZmbDg8dooHkMn90NWF3hdWEwrfL3XGVmS4F7gWtruOzlZpZnZnmbNm2KIVLD0qVNM341biAzVmzjyfeXhx1HROqY6rYgTgV+B2QS2fVzf/C4Hri1tgK4+6Pu3hu4Cfh5DZd9wt1z3T03IyOjtiLVK+cM68YpAzpx3xuLWbJhZ9hxRKQOqW4016fcfQxwqbuPqfAY5+6TY/jsNUBWhdeZwbSqPAucfYjLShXMjLvOGUTLJmncMHE2pWW6W6yIxKa6XUwXB097mtn1lR8xfPYMoK+ZZZtZYyIHnadU+o6+FV5+g8j9Jgjmu8DMmphZNtAX0LUXhyijVRN+ffZRzCncwR+nLw07jojUEdWdxXTgtJeWh/LB7l5qZlcDrwOpwJPuPt/M7gDy3H0KcLWZnQTsB7YBlwTLzjez54AFRK7avkpnMB2eMwZ14ayhXfn91CWMPbIjA7vqUhYRqZ7VlwupcnNzPS8vL+wYSW37nhJOefBd2rdozItXj6RJWmrYkUQkZGaW7+650d6L5UrqDDO71cyeMLMnDzxqP6bEW9vmjbnnvMEsWr+T37+15OALiEiDFsuFci8C7xG59ah289RxY/p35Fu5WTz+zlJOGtCJ4d3bhR1JRJJULAXR3N1vinsSSZifnxkZhuMnz83mP9ceT7PG2tUkIv8tliupXzazM+KeRBKmVdNG3Dd+MMs27+be1xeFHUdEklQsBXEdkZLYa2ZFZrbTzIoOupQktRF9OnDpiJ789YMVfLh0c9hxRCQJHbQg3L2Vu6e4ezN3bx28bp2IcBJfN53Wn+wOLbhx0hx27dM9oETkq2I5i2l4lEdvM4vl+IUksWaNU/ndhMGs3b6Xu/6zMOw4IpJkYtnF9BjwMfDn4PExMBFYbGanxDGbJEBOj/ZcPqo3z3y6iumLN4YdR0SSSCwFsRYY5u457p4DDAWWEbnPw73xDCeJ8b8n96Vfp5bc9PwcduzZH3YcEUkSsRREP3eff+BFcMOf/u6+LH6xJJGapKXywPlD2bKrhNtfmn/wBUSkQYilIOab2R/N7ITg8RiwILirnP65WU8c1a0N14ztywufreG1eevCjiMiSSCWgriUyB3dfhw8lgXT9gNj4hVMEu/KMb0Z1K0NP3thHpt37Qs7joiELJbTXPe6+/3ufk7w+J2773H3cnfflYiQkhiNUlO4//wh7Cwu5ecvzKO+DOQoIocmltNc+5rZJDNbYGbLDjwSEU4Sr1+nVtxwSj9em7+eF2etDTuOiIQoll1MfwX+SOS+DGOAvwH/iGcoCdcPju9FTo92/OLFeazfURx2HBEJSSwF0czdpxK5d8RKd7+dyN3fpJ5KTTHunzCE/WXOzZPnaFeTSAMVS0HsM7MUYImZXW1m53CId5mTuqNnhxbcckZ/pi/exL9mrA47joiEINbB+poD1wI5wHcIbg0q9dvFx/RgZJ907nx5Aau37gk7jogkWCxnMc1w913uXujul7n7ue7+cSLCSbhSUox7xw/BzPjppNmUl2tXk0hDUuWAe2Y2pboF3X1c7ceRZNOtbTN+8c0B3DhpDk99tILLRmaHHUlEEqS6EVmPA1YDzwCfAJaQRJJ0JuRk8vq89dz96iJG9cugd4YOQYk0BNXtYuoM3AocBfyeyOB8m939HXd/JxHhJDmYGb89dxBNG6Xyk4mzKS0rDzuSiCRAlQXh7mXu/pq7XwIcS2S4jelmdnXC0knS6Ni6KXeefRSfrdrOE+/pOkmRhqDag9Rm1sTMziVyYdxVwMPAC4kIJsnnm4O78I1BXXjwzc/5aOmWsOOISJxVd5D6b0R2L70C/Mrd5yUslSQlMwu2IrZx4Z8/5vi+HbhmbF+Ozm4fdjQRiQOr6ipZMysHdgcvK85kgCfbfalzc3M9Ly8v7BgNwu59pTz9yUqeeHc5m3ft45js9lx3Yl+O652Omc5lEKlLzCzf3XOjvldfhlFQQSRe8f4ynvl0FY+/s5QNRfvI6dGOa8b24YR+GSoKkTpCBSFxVby/jIn5hTw+fSlrtu9lSGbk5kMnHtlRRSGS5FQQkhAlpeVMnlnIY9OXsmrrHgZ0ac01Y/tw6sDOpKSoKESSkQpCEqq0rJwXZ63l0bcLWLZ5N/06teTqsX35xqAupKooRJJKdQURyw2DWgSjuWJm/cxsnJk1qu2QUn+kpaZwXk4mb15/Ag9fOAx3uPaZzzj5wXd4Pr9QF9qJ1BEH3YIws3zgeKAd8AEwAyhx94viHy922oJIXuXlzuvz1/PwtAIWriuie/vmXDm6N+cOz6RxWiwDCotIvBzWFgSREtkDnAs85u4TgIG1GVDqt5QU4/RBXXjl2q/z5+/m0rZ5I26ePJcxv5vO3z9eyb7SsrAjikgUMRWEmR0HXAT8J5iWGsuHm9lpZrbYzArM7OYo718f3Ot6jplNNbMeFd4rM7NZwaPakWWlbjAzTh7QiRevGsn/XfY1Ordpym3/nseoe9/mrx8sp3i/ikIkmcSyi+kE4AbgA3e/x8x6AT9292sPslwq8DmRQf4KieyautDdF1SYZwzwibvvMbMfAaPd/VvBe7vcPeZhQ7WLqe5xdz5auoXfT13CJ8u30qFlEy4flc1Fx/SgRZPqBhoWkdpSa2cxBQerW7p7UQzzHgfc7u6nBq9vAXD331Yx/zDgEXcfGbxWQTQgny7fyh+mLeG9JZtp17wRPzi+F989rgetmup8CJF4OtyzmP5pZq3NrAUwD1hgZj+N4Xu7EbmfxAGFwbSqfB94tcLrpmaWZ2Yfm9nZVWS7PJgnb9OmTTFEkmR1dHZ7/v79Y5h85QiGZrXlvtcXM/LuaTz01ufs2LM/7HgiDVIsxyAGBFsMZxP5A55N5L7UtcbMLgZygfsqTO4RtNq3gYfMrHfl5dz9CXfPdffcjIyM2owkIRnevR1/vexoXrr66xzbK52H3lrC1++Zxn2vL2Lr7pKw44k0KLEURKPguoezgSnuvp+vDt5XlTVAVoXXmcG0rzCzk4CfAePcfd+B6e6+JvjvMmA6MCyG75R6YlBmG574bi6vXnc8o/pl8Nj0pXz9nmn89pWFbNq57+AfICKHLZaC+BOwAmgBvBucaXTQYxBEDkr3NbNsM2sMXAB85Wyk4LjDn4iUw8YK09uZWZPgeQdgJLAAaXCO7NKaRy8azhs/HsUpAzrx5/eWcfy90/jVS/PZUFQcdjyReu2QhtowszR3L41hvjOAh4icFvuku99lZncAee4+xczeAgYB64JFVrn7ODMbQaQ4yomU2EPu/v+q+y4dpG4Ylm/ezaNvF/DCZ2tINeNbX8viitG96da2WdjRROqkwzqLyczaAL8ERgWT3gHucPcdtZryMKkgGpbVW/fw2PSlTMqPnAdx3vBMrhzdh+7pzUNOJlK3HG5BPE/k7KWngknfAYa4+7m1mvIwqSAaprXb9/L4O0t5dsZqysqds4d246oxvemVEfMZ0iIN2uEWxCx3H3qwaWFTQTRsG4qKeeLdZTz9yUpKSss5c3BXrh7bh36dWoUdTSSpHe5YTHvN7OsVPmwksLe2wonUhk6tm3LbmQN4/6ax/HBUL95auIFTH3qXK5/OZ8HaWM6pEJHKYtmCGAL8DWgTTNoGXOLuc+KcrUa0BSEVbdtdwpMfLOf/PljBzn2lnHRkJ649sQ+DM9uGHU0kqdTKUBtm1hrA3YvM7Mfu/lAtZjxsKgiJZsfe/fzfByt48oPl7Ni7n9FHZHDN2L7k9GgXdjSRpFDrd5Qzs1Xu3v2wk9UiFYRUZ2fxfv7+8Ur+8t5ytu4uYWSfdK4Z25dje6WHHU0kVPEoiNXunnXwORNHBSGx2FNSytMfr+JP7y5j8659HN2zPdee2JeRfdIx0+1QpeHRFoRIJcX7y3j201U8/s4y1hcVM6x7W64d25fRR2SoKKRBOaSCMLOdRB9zyYBm7p5UA/arIORQ7CstY2JeIX+cvpQ12/cyqFsbrh7bh5OP7ERKiopC6r9a34JIRioIORz7y8p5YeYaHnm7gFVb99C/cyuuGduX04/qrKKQek0FIRKj0rJypsxeyyNvF7Bs0276dGzJNWP7cObgrqSqKKQeUkGI1FBZufPK3HU8Mq2AxRt2kt2hBVeO7s3Zw7rRKDWW60tF6gYVhMghKi933liwnoenFrBgXRGZ7Zpx5eg+jM/JpHGaikLqPhWEyGFyd6Yt2sjDU5cwu3AHXds05YrRvTk/N4umjVLDjidyyFQQIrXE3Xl3yWb+MHUJeSu30bFVEy4f1YuLjulBs8YqCql7VBAitczd+WjZFh6euoSPl20lvUVjfjiqFxcf24OWTZLqDHCRaqkgROJoxoqtPDx1Ce8t2Uzb5o34/shsLhnZk9ZNG4UdTeSgVBAiCfDZqm08Mq2AqYs20qppGpeNzOZ7I3vStnnjsKOJVEkFIZJA89bs4A/TlvD6/A20aJzKd0f05LKRPenYqmnY0UT+iwpCJASL1hfxh2kFvDJ3HSlmjO6XwYTcTMb276RTZCVpVFcQOpomEif9O7fm0W8PZ9mmXTyXV8jkmYVMXbSR9i0ac9bQrkzIyWJA19ZhxxSpkrYgRBKktKyc9wo2MymvkDcXbKCkrJyBXVszISeTs4Z2o10LHauQxNMuJpEks213CVNmr2Vi/mrmrSmiUapx0pGdmJCbyai+GaRpOA9JEBWESBJbuK6IiXmF/HvWGrbuLqFjqyacM7wbE3Ky6NOxZdjxpJ5TQYjUASWl5UxbtJFJ+at5e/EmysqdYd3bMiEnizOHdNF1FRIXKgiROmbTzn38+7M1TMxfzecbdtEkLYXTjurMhJwsRvRO1z0qpNaoIETqKHdnTuEOJuavZsqstRQVl9KtbTPOG96N8TlZdE9vHnZEqeNUECL1QPH+Mt5YsIFJ+YW8t2QT7nBMdnsm5GZx+lGdaaExoOQQqCBE6pl1O/YyeeYaJuatZsWWPbRonMoZg7owITeLr/Vsh5l2QUlsVBAi9ZS7k7dyGxPzVvOfOevYXVJGz/TmjM/J5NzhmXRt2yzsiJLkVBAiDcCeklJembueiXmr+WT5Vszg6306MCE3i1MGdNKNjSQqFYRIA7Nqyx4mzSzk+fxC1mzfS+umaXxzSFcm5GYxJLONdkHJF1QQIg1UeXnkxkYT81bz6rz17Cstp1+nlozPyeTsYd00wqyEVxBmdhrweyAV+Iu7313p/euBHwClwCbge+6+MnjvEuDnway/dvenqvsuFYRI9YqK9/Py7HVMzF/NZ6u2k5pijDkig/E5WYzt31EjzDZQoRSEmaUCnwMnA4XADOBCd19QYZ4xwCfuvsfMfgSMdvdvmVl7IA/IBRzIB3LcfVtV36eCEIldwcadTMpfw+SZhWzcuY/2LRpz9tBujM/J1AizDUxYBXEccLu7nxq8vgXA3X9bxfzDgEfcfaSZXUikLP4neO9PwHR3f6aq71NBiNRcaVk57y3ZzMT81by5YAP7y1wjzDYwYd0PohuwusLrQuCYaub/PvBqNct2q9V0IkJaagpj+ndkTP+ObNtdwouz1jAxv5DbX1rAb15ZxEkDOjIhJ4vj+3bQCLMNUFJcemlmFxPZnXRCDZe7HLgcoHv37nFIJtJwtGvRmEtHZnPpyGwWrC1iYv5qXpy1llfmrtcIsw1UPAtiDZBV4XVmMO0rzOwk4GfACe6+r8KyoystO73ysu7+BPAERHYx1UZoEYEBXVvzy64DueX0I78YYfYv7y3nT+8s0wizDUg8j0GkETlIfSKRP/gzgG+7+/wK8wwDJgGnufuSCtPbEzkwPTyYNJPIQeqtVX2fjkGIxNfGncWREWbzClmycRdNG6Vw2sDOTMjN4rheGmG2rgrzNNczgIeInOb6pLvfZWZ3AHnuPsXM3gIGAeuCRVa5+7hg2e8BtwbT73L3v1b3XSoIkcRwd2YX7mBi3mqmzF7LzgMjzOZkMiEnk6z2GmG2LtGFciISFwdGmJ2Yt5r3CzbjDsf2as/4nCzOGNSZ5o2T4jCnVEMFISJxt3b7XibPLGRSfuEXI8x+Y3BkhNncHhphNlmpIEQkYdydGSuCEWbnrmNPSRnZHVoEI8x2o0sbjTCbTFQQIhKK3ftKeXWeRphNZioIEQndyi27eT6/kOdnrvlihNlxQ7syISeLwRphNjQqCBFJGuXlzodLtzAxfzWvVRhhdkJOFmcP60ZGqyZhR2xQVBAikpR27N3Py3PWMjGvkFmrvxxh9szBXRnRO52OrTUcebypIEQk6RVs3MnE/EImz1zDpp2RQRX6dGzJiN7pjOidzrG90mnbXIMH1jYVhIjUGWXlzsJ1RXy4dDMfLt3Cp8u3sqekDDMY0KV1UBgd+Fp2e1o20XUWh0sFISJ11v6ycuYUbufDgi18uHQL+au2UVJaTmqKMSSzDSN6d2BE73SG92ins6IOgQpCROqN4v1lzFy5jQ+XbuGDpZuZU7iDsnKncVoKuT3aMaJ3Osf17sDgzDY00hDlB6WCEJF6a2fxfmas2PrFFsaCdUUAtGicytHZ7RnRuwPH9U5nQJfWGlAwirBuGCQiEnetmjZibP9OjO3fCYCtu0v4ZFmkLD5cupm3Fy8EoG3zRqbV+lsAAAiZSURBVBybnc6IPpGD3r0zWurai4NQQYhIvdK+RWNOH9SF0wd1AWBDUTEfBWXxQcEWXpu/HoCMVk0Y0TudkcEWhkah/W/axSQiDYa7s3rr3i/OkPpw6RY274qcUpvVvhkjenVgRJ90juvVcK7B0DEIEZEo3J2Cjbu+2B310dItFBWXAg3nGgwVhIhIDBriNRgqCBGRQ1DxGowPlm5m5srtlJSVk5ZiDMlqG5xSm87w7nX3GgwVhIhILSjeX0b+ym1fbGHUh2swVBAiInFQH67B0HUQIiJxUNNrMEb2iWxh9M5oUSeuwVBBiIjUklivwegYXIMxIsmvwdAuJhGRBEjWazB0DEJEJMkkyzUYKggRkSRX3TUYA7u2/mJ31Nd61u41GCoIEZE6JlHXYKggRETquOquwThlQCce+fbwQ/pcneYqIlLHNW2Uysg+HRjZpwPw1WswmjSKz0V5KggRkTqo8jUY8VA3rgUXEZGEU0GIiEhUKggREYlKBSEiIlGpIEREJCoVhIiIRKWCEBGRqFQQIiISVb0ZasPMNgErD+MjOgCbaylObVKumlGumlGumqmPuXq4e0a0N+pNQRwuM8urajySMClXzShXzShXzTS0XNrFJCIiUakgREQkKhXEl54IO0AVlKtmlKtmlKtmGlQuHYMQEZGotAUhIiJRqSBERCSqBlUQZnaamS02swIzuznK+03M7F/B+5+YWc8kyXWpmW0ys1nB4wcJyvWkmW00s3lVvG9m9nCQe46ZHdo9D2s/12gz21Fhff0iQbmyzOxtM1tgZvPN7Loo8yR8ncWYK+HrzMyamtmnZjY7yPWrKPMk/DcZY65QfpPBd6ea2Wdm9nKU92p3fbl7g3gAqcBSoBfQGJgNDKg0z5XA48HzC4B/JUmuS4FHQlhno4DhwLwq3j8DeBUw4FjgkyTJNRp4OYT11QUYHjxvBXwe5X/LhK+zGHMlfJ0F66Bl8LwR8AlwbKV5wvhNxpIrlN9k8N3XA/+M9r9Xba+vhrQFcTRQ4O7L3L0EeBY4q9I8ZwFPBc8nASeamSVBrlC4+7vA1mpmOQv4m0d8DLQ1sy5JkCsU7r7O3WcGz3cCC4FulWZL+DqLMVfCBetgV/CyUfCofNZMwn+TMeYKhZllAt8A/lLFLLW6vhpSQXQDVld4Xch//0i+mMfdS4EdQHoS5AI4L9glMcnMsuKcKVaxZg/DccEuglfNbGCivzzYtB9G5F+fFYW6zqrJBSGss2B3ySxgI/Cmu1e5vhL4m4wlF4Tzm3wIuBEor+L9Wl1fDakg6rKXgJ7uPhh4ky//hSDRzSQyvswQ4A/AvxP55WbWEnge+LG7FyXyu6tzkFyhrDN3L3P3oUAmcLSZHZWI7z2YGHIl/DdpZmcCG909P97fdUBDKog1QMWWzwymRZ3HzNKANsCWsHO5+xZ33xe8/AuQE+dMsYplnSacuxcd2EXg7q8AjcysQyK+28waEfkj/LS7T44ySyjr7GC5wlxnwXduB94GTqv0Vhi/yYPmCuk3ORIYZ2YriOyKHmtm/6g0T62ur4ZUEDOAvmaWbWaNiRzAmVJpninAJcHz8cA0D472hJmr0j7qcUT2ISeDKcB3gzNzjgV2uPu6sEOZWecD+13N7Ggi/z+P+x+V4Dv/H7DQ3R+oYraEr7NYcoWxzswsw8zaBs+bAScDiyrNlvDfZCy5wvhNuvst7p7p7j2J/J2Y5u4XV5qtVtdX2qEuWNe4e6mZXQ28TuTMoSfdfb6Z3QHkufsUIj+iv5tZAZGDoBckSa5rzWwcUBrkujTeuQDM7BkiZ7d0MLNC4JdEDtjh7o8DrxA5K6cA2ANcliS5xgM/MrNSYC9wQQKKHiL/wvsOMDfYfw1wK9C9QrYw1lksucJYZ12Ap8wslUghPefuL4f9m4wxVyi/yWjiub401IaIiETVkHYxiYhIDaggREQkKhWEiIhEpYIQEZGoVBAiIhKVCkLqJTMrqzDS5iyLMkruYXx2T6tiJNlK891uZnvMrGOFabuqW6a2M4gcjgZzHYQ0OHuDoRLCthm4Abgp7CAVmVlaMFaPSJW0BSENipmtMLN7zWyuRcb87xNM72lm04LB16aaWfdgeiczeyEYxG62mY0IPirVzP5skfsFvBFccRvNk8C3zKx9pRxf2QIws5+Y2e3B8+lm9qCZ5ZnZQjP7mplNNrMlZvbrCh+TZmZPB/NMMrPmwfI5ZvaOmeWb2esHrvoNPvchM8sD/uueECKVqSCkvmpWaRfTtyq8t8PdBwGPEBkdEyID1D0VDL72NPBwMP1h4J1gELvhwPxgel/gUXcfCGwHzqsixy4iJVHTP8gl7p4LPA68CFwFHAVcamYHRuc8AnjM3Y8EioArLTLm0h+A8e6eE3z3XRU+t7G757r7/TXMIw2QdjFJfVXdLqZnKvz3weD5ccC5wfO/A/cGz8cC34XICJ/ADjNrByx39wPDVuQDPavJ8jAwy8x+V4P8B8bjmgvMPzBek5ktIzIY23Zgtbt/EMz3D+Ba4DUiRfJmMLRSKlBxrKd/1SCDNHAqCGmIvIrnNbGvwvMyoKpdTLj7djP7J5GtgANK+eoWfNMqPr+80neV8+XvtnJ2J3I3tPnuflwVcXZXlVOkMu1ikoboWxX++1Hw/EO+HNjsIuC94PlU4EfwxU1k2hzidz4A/A9f/nHfAHQ0s3QzawKceQif2d3MDhTBt4H3gcVAxoHpZtbIQrhhktQPKgipryofg7i7wnvtzGwOkeMC/xtMuwa4LJj+Hb48ZnAdMMbM5hLZlTTgUMK4+2bgBaBJ8Ho/cAfwKZEbzlQe5joWi4GrzGwh0A74Y3Db2vHAPWY2G5gFjKjmM0SqpNFcpUGxyM1WcoM/2CJSDW1BiIhIVNqCEBGRqLQFISIiUakgREQkKhWEiIhEpYIQEZGoVBAiIhLV/wcTPEnKAHwn2wAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xyzMv8cWHNiB",
        "colab_type": "text"
      },
      "source": [
        "## Result"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TjCBt7nTGCFL",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 71
        },
        "outputId": "a2a00db6-4c51-43ce-e396-5c080997faff"
      },
      "source": [
        "test_loss, test_accuracy = model.evaluate(test_ds, steps=math.ceil(md.splits[\"test\"].num_examples/32))\n",
        "print('accuracy:', test_accuracy*100, \"%\")"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "313/313 [==============================] - 1s 3ms/step - loss: 0.2456 - accuracy: 0.9149\n",
            "accuracy: 91.4900004863739 %\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "y9MNtji5GGWU",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 316
        },
        "outputId": "6458d94f-559b-4590-de96-2cd700d99f51"
      },
      "source": [
        "for test_images, test_labels in test_ds.take(1):\n",
        "\n",
        "    test_images_n = test_images.numpy()\n",
        "    test_labels_n = test_labels.numpy()\n",
        "    predictions = model.predict(test_images_n)\n",
        "\n",
        "    i = int(input(\"index? (0-31)\"))\n",
        "    \n",
        "    plt.imshow(test_images[i].numpy().reshape((28,28)))\n",
        "    print(\"labels: \", test_labels_n[i])\n",
        "    print(\"prediction: \", np.argmax(predictions[i]))"
      ],
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "index? (0-31)1\n",
            "labels:  4\n",
            "prediction:  4\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}