CNN-Classification.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Vdq9nxUA8WvA"
      },
      "outputs": [],
      "source": [
        "# 1. Thêm các thư viện cần thiết\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Dropout, Activation, Flatten\n",
        "from keras.layers import Conv2D, MaxPooling2D\n",
        "from keras.utils import np_utils\n",
        "from keras.datasets import mnist"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# 2. Load dữ liệu MNIST\n",
        "(X_train, y_train), (X_test, y_test) = mnist.load_data()\n",
        "X_val, y_val = X_train[50000:60000,:], y_train[50000:60000]\n",
        "X_train, y_train = X_train[:50000,:], y_train[:50000]\n",
        "print(X_train.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "aIJ4VZU-8m3I",
        "outputId": "14c036d1-edec-496f-b81c-32a4e8e0aec9"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz\n",
            "11490434/11490434 [==============================] - 0s 0us/step\n",
            "(50000, 28, 28)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# 3. Reshape lại dữ liệu cho đúng kích thước mà keras yêu cầu\n",
        "X_train = X_train.reshape(X_train.shape[0], 28, 28, 1)\n",
        "X_val = X_val.reshape(X_val.shape[0], 28, 28, 1)\n",
        "X_test = X_test.reshape(X_test.shape[0], 28, 28, 1)"
      ],
      "metadata": {
        "id": "gw8RnHzp8oTC"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 4. One hot encoding label (Y)\n",
        "Y_train = np_utils.to_categorical(y_train, 10)\n",
        "Y_val = np_utils.to_categorical(y_val, 10)\n",
        "Y_test = np_utils.to_categorical(y_test, 10)\n",
        "print('Dữ liệu y ban đầu ', y_train[0])\n",
        "print('Dữ liệu y sau one-hot encoding ',Y_train[0])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2tPpFXlb_F8j",
        "outputId": "043abc28-4e9b-4445-b641-9e87a5da26c4"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Dữ liệu y ban đầu  5\n",
            "Dữ liệu y sau one-hot encoding  [0. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# 5. Định nghĩa model\n",
        "model = Sequential()\n",
        "\n",
        "# Thêm Convolutional layer với 32 kernel, kích thước kernel 3*3\n",
        "# dùng hàm sigmoid làm activation và chỉ rõ input_shape cho layer đầu tiên\n",
        "model.add(Conv2D(32, (3, 3), activation='sigmoid', input_shape=(28,28,1)))\n",
        "\n",
        "# Thêm Convolutional layer\n",
        "model.add(Conv2D(32, (3, 3), activation='sigmoid'))\n",
        "\n",
        "# Thêm Max pooling layer\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "\n",
        "# Flatten layer chuyển từ tensor sang vector\n",
        "model.add(Flatten())\n",
        "\n",
        "# Thêm Fully Connected layer với 128 nodes và dùng hàm sigmoid\n",
        "model.add(Dense(128, activation='sigmoid'))\n",
        "\n",
        "# Output layer với 10 node và dùng softmax function để chuyển sang xác xuất.\n",
        "model.add(Dense(10, activation='softmax'))\n",
        "# Compile model, chỉ rõ hàm loss_function nào được sử dụng, phương thức đùng để tối ưu hàm loss function.\n",
        "model.compile(loss='categorical_crossentropy',\n",
        "              optimizer='adam',\n",
        "              metrics=['accuracy'])"
      ],
      "metadata": {
        "id": "bVoNpF4f_U76"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 7. Thực hiện train model với data\n",
        "H = model.fit(X_train, Y_train, validation_data=(X_val, Y_val),\n",
        "          batch_size=32, epochs=10, verbose=1)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "nuz6AFeR_pnD",
        "outputId": "00c64a31-ffc8-475a-b34f-89d840b810e6"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/10\n",
            "1563/1563 [==============================] - 105s 66ms/step - loss: 0.2852 - accuracy: 0.9200 - val_loss: 0.0654 - val_accuracy: 0.9832\n",
            "Epoch 2/10\n",
            "1563/1563 [==============================] - 99s 63ms/step - loss: 0.0563 - accuracy: 0.9840 - val_loss: 0.0577 - val_accuracy: 0.9824\n",
            "Epoch 3/10\n",
            "1563/1563 [==============================] - 97s 62ms/step - loss: 0.0352 - accuracy: 0.9900 - val_loss: 0.0440 - val_accuracy: 0.9871\n",
            "Epoch 4/10\n",
            "1563/1563 [==============================] - 99s 63ms/step - loss: 0.0235 - accuracy: 0.9930 - val_loss: 0.0412 - val_accuracy: 0.9880\n",
            "Epoch 5/10\n",
            "1563/1563 [==============================] - 98s 63ms/step - loss: 0.0150 - accuracy: 0.9961 - val_loss: 0.0433 - val_accuracy: 0.9865\n",
            "Epoch 6/10\n",
            "1563/1563 [==============================] - 98s 63ms/step - loss: 0.0105 - accuracy: 0.9973 - val_loss: 0.0410 - val_accuracy: 0.9885\n",
            "Epoch 7/10\n",
            "1563/1563 [==============================] - 97s 62ms/step - loss: 0.0072 - accuracy: 0.9982 - val_loss: 0.0356 - val_accuracy: 0.9901\n",
            "Epoch 8/10\n",
            "1563/1563 [==============================] - 99s 63ms/step - loss: 0.0050 - accuracy: 0.9988 - val_loss: 0.0447 - val_accuracy: 0.9882\n",
            "Epoch 9/10\n",
            "1563/1563 [==============================] - 98s 62ms/step - loss: 0.0031 - accuracy: 0.9994 - val_loss: 0.0421 - val_accuracy: 0.9897\n",
            "Epoch 10/10\n",
            "1563/1563 [==============================] - 105s 68ms/step - loss: 0.0039 - accuracy: 0.9991 - val_loss: 0.0514 - val_accuracy: 0.9871\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#Vẽ đồ thị loss, accuracy của traning set và validation set\n",
        "fig = plt.figure()\n",
        "numOfEpoch = 10\n",
        "plt.plot(np.arange(0, numOfEpoch), H.history['loss'], label='training loss')\n",
        "plt.plot(np.arange(0, numOfEpoch), H.history['val_loss'], label='validation loss')\n",
        "plt.plot(np.arange(0, numOfEpoch), H.history['accuracy'], label='accuracy')\n",
        "plt.plot(np.arange(0, numOfEpoch), H.history['val_accuracy'], label='validation accuracy')\n",
        "plt.title('Accuracy and Loss')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Loss|Accuracy')\n",
        "plt.legend()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 312
        },
        "id": "Z--P1lzHKzNk",
        "outputId": "1fab6684-f598-448c-f814-8ee622236095"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f4a7230abe0>"
            ]
          },
          "metadata": {},
          "execution_count": 7
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEWCAYAAABrDZDcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3deXwU9f348dd7jyQkQAiXymVoPYCEQMKpVEVRi1rBC/EW6vHVetT69UB/KtbWfm21lHq2aL1aFSkWtRalUqFoi5ZDQC4rAsohpxACOXf3/ftjZjebZBOykM0S9v2EeezMZz4z855JMu+59jOiqhhjjEldnmQHYIwxJrksERhjTIqzRGCMMSnOEoExxqQ4SwTGGJPiLBEYY0yKs0RgTAshIutF5PRkx2EOP5YITNKIyFwR2SUi6cmOpaUTkRdF5OfJjsO0TJYITFKISC5wEqDAqGZetq85l2fMoc4SgUmWq4CPgReBq6NHiEh3EfmLiGwXkZ0i8mTUuOtEZJWIlIjIShEpcstVRI6Jqhc5QhaR4SKyUUTuFpEtwAsikiMi77jL2OX2d4uavr2IvCAim93xb7rly0Xk3Kh6fhHZISKFtVewEcuYKyI/E5F/uevzdxHpGDX+ShH5yt0G/+9AN7S7zdaIyLci8raIdHHLRUR+IyLbRGSPiHwmIvnuuLPd7VsiIptE5I4DXb459FkiMMlyFfCK231fRI4AEBEv8A7wFZALdAWmuuPGAA+607bFOZPY2cjlHQm0B44Grsf53X/BHe4BlAFPRtX/I5AJ5AGdgd+45S8DV0TVOxv4RlU/jbHM/S0D4DJgvLuMNOAOd137AM8AVwJdgA5AN+IkIqcB/wdcDByFs12nuqPPBE4GjgOy3Trh7fkH4H9UtQ2QD3wQ77JNC6Kq1lnXrB3wPaAK6OgOrwZ+4vafAGwHfDGmmwX8uJ55KnBM1PCLwM/d/uFAJZDRQEz9gV1u/1FACMiJUa8LUAK0dYenA3c1cr0jy3CH5wL3RQ3/CHjP7X8AmBo1Lstdh9PrmXdkfWuV/wH4VdRwa3fb5wKnAf8FhgKeWtN9DfxPeD2tO7w7OyMwyXA18HdV3eEOv0r15aHuwFeqGogxXXfgywNc5nZVLQ8PiEimiPzevfSyB5gHtHPPSLoD36rqrtozUdXNwL+AC0WkHXAWzllNHftZRtiWqP5SnB01OAlnQ9Ry99H4s59oXXDOAsLz2evOp6uqfoBzhvIUsE1EpohIW7fqhThnO1+JyD9F5IQDWLZpISwRmGYlIq1wLkGcIiJb3Gv2PwH6iUg/nJ1fj3pu6G4AvlvPrEtxLuWEHVlrfO1mdv8XOB4YoqptcS6RAIi7nPbujj6Wl3AuD40B5qvqpnrqNbSM/fkGJyE5E4hk4lweitdmnEtT4flkufPZBKCqj6vqAKAPziWiO93yBao6GueS1ZvAtANYtmkhLBGY5nYeEMTZ8fR3u97AhzjX/v+DsxN8RESyRCRDRIa50z4H3CEiA9wbnceISHgntwS4TES8IjISOGU/cbTBuWa/W0TaAxPDI1T1G+Bd4Gn3hq9fRE6OmvZNoAj4Mc49g7iX0QjTgR+IyPdEJA14iP3/vXrd7RXu0oDXgPEi0l+cx3R/AXyiqutFZJCIDBERP7APKAdCIpImIpeLSLaqVgF7cC6VmcOUJQLT3K4GXlDVr1V1S7jDuURxOc7R8rnAMTjXqTcCYwFU9c/AwziXkkpwdsjt3fn+2J1utzufN/cTx2SgFbAD5+ml92qNvxLnWvpqYBtwW3iEqpYBbwA9gb8cxDLqpaorgJtw1vUbYBfOtmjIBJzEE+4+UNXZwP1uvN/gnFFd4tZvCzzrzvsrnEtGj7rjrgTWu5e0bsDZpuYwJar2Yhpj4iUiDwDHqeoV+61szCHOvlhjTJzcyzzX4Bw1G9Pi2aUhY+IgItfh3Ex+V1XnJTseY5qCXRoyxpgUZ2cExhiT4lrkPYKOHTtqbm5ussMwxpgWY9GiRTtUtVOscS0yEeTm5rJw4cJkh2GMMS2GiHxV3zi7NGSMMSnOEoExxqQ4SwTGGJPiLBEYY0yKs0RgjDEpLqGJQESed1+Dt7ye8SIij7uv0Vsm7msHjTHGNJ9EnxG8CIxsYPxZwLFudz3Oq/mMMcY0o4R+j0BV54lIbgNVRgMvq9POxcci0k5EjnLbgzemRVNVQhpCcV4HGCLkDLvlIUKRVwXWHheeJvwvpCFQqqeJGl+nvjs/lOpxtevvZx4hDUXWAagTj/O/eh6ROtHjoupHN2VTO9ZwWXh5deZba/7hOtHbuaF51V5mrOmi60WvW/Tyw0Sc9woJNT+jXzdUe1ydaaTuu4nqrRtV3srXiouPv7jOtAcr2V8o60rU6/hw2lvvitNueg0icj3OWQM9evRoluBMfFSVgAYIhoIEQgGCGqQqVBXpD4SccVWhquph97MqVFVjukiZ2x9dHumilhXQQJ16seYZXTcYCtaYLqShyA4x0u/uVKN3kNHl4R1mfeXm8BDeGUcnkmTokNHhsEwEjaaqU4ApAAMHDkzplvKqQlVUBCqoCDpdebA8MhzuLw+WO8OB8up6gXIqAuVUVJQSKC8jUFlOMFhFKBQkGHR2kKFggGAoQDAYIhQMENIggWAVIQ0SDAXRoPMZDAUi06mGCIUChELOzs+jgLqfgKjb4XbusFPX6QmJoAIhIfIZ7o8Me9xhnP6QgNfjQzwePD4f4vHi9fjweL14vH68Xh/i9eL1+txyH16vH4/XSytfK3weHz7xOZ8eHyKCV7x4xIMgeFTwqiCqeN1+T6hmv4Rw+lXxhASvu25eBQnhljt1JKR4FDwhZ9tIyBknWl0uIXW2T8h5obi448PTElJ32C13x4WnEdXqOm4Z4TrR/VH1CEWVqUIw5PSLgMeLeD3g8YDX4wx7nHLccvF6nfEeD+LxgNeLuOPFU11HItO48xS33OtBPF5nPpFlRE2jChqCkEIoFBUzaCgYGUad8YRCkfUMT6PqThcMufWq1zUyb1UIBmuOD4XQkFMmXh/i84L7e4XPjdnjBZ8H8frA63XWP3p8uN/j9ns94PM528frA5/X3UY+1OdsM/VWTys+b2RYfInZZSc7EWwi6r2sQDe3LCFK5s51fsACiDi/pOLunkScj4bK3HLxSPVwuK5HnFO6Wl3NMg87y3cyf8vH7C3dTaC8jGBFOYGKMkIVFQQryglVVhCqrESjOqoCSFUVVAaQQBBfIIQvCP4A+N1PXxD8QY3qhzYByHH7/UHwBSAtmKitmywHsULhnZaI0w8QDKKhkLNDaKnc9QnvhKPXMVYZXg8itco8nsiOMbw91N3J1hiOKicYdA4GWtq2cxNZnW0UVYaIs65uRyDgfIaa98zP27EjbT76sMnnm+xE8DZws4hMBYYAxYm8P7Dptp+g5eWJmn2j9TnA6UJeDyG/F/X7Ub8P9fvA74c0P5KehqSl4UlLw5OWjic9HW9aOt70DLzpGfjSW+FLb4XHrSf+NMTvd45OIgnLEyP5xSirk/jqm466CTfWdOAefQXRkHP05/yRuf2hUOTILbLTCYWcuuEjtvrqRvpr1Y01neIe/Xr3+xk+mmt4vJfoI+bYn/Uto9YOKUZ/zJ16eOeVZJEkEU4mwZDz8w3vPGMllHCd8Pjw74BHqtdVxNlu4nHPTKLPUsJnGVF1Y22f2jv6pljPUCiSHKIThQaCEAz3B6rXs3ZZIIgGq/ud8aEaZRoM4ElLa6KfUE0JTQQi8howHOgoIhtxXt7tB1DV3wEzgbOBNUApMD6R8eS++orzQ8A97VON7ABilqlGyuuUhZzrwc401ClTVb4qXs/CbxawdNsSygNl5KS1Y0DnIoo6F9Kh9RF4MzKqd8pRO3KJ1fn9zi+wMS1AJFklO5AEq7Ge6enJDueAJfqpoUv3M15xXtDdLDL6HOixeONt3beVd9a+w1tfvsW6snVkdMjgjKIzueyY0Qw6chAesZ25MebQkuxLQ4eFimAFc76ew5tfvsn8zfMJaYiizkWMO3EcZx59Jq3TWic7RGOMqZclggOkqizfsZw317zJu+vfpaSyhCOzjuTavtcy+ruj6dHWHnE1xrQMlgjitK10m3PpZ81brC1eS4Y3gxFHj+C8Y85j8JGD7dKPMabFsUTQCBXBCuZsmMNba97i35v/TUhDFHYu5METHuTM3DNpk9Ym2SEaY8wBs0RQD1Vlxc4VzqWfde+yp3IPR2QewTX51zDqu6PIzc5NdojGGNMkLBHUsr10e+TSz5fFX5LuTWdEjxGMPmY0Q44cgtfjTXaIxhjTpCwRAJXBSuZumMuba97kX5v/RUhD9OvUj4knTOT7ud+3Sz/GmMNayiYCVWXlzpW8ueZNZq6byZ7KPXTO7MwP83/IqO+Oomd2z2SHaIwxzSLlEsGOsh2886Xzha81u9eQ5kljRA/nqZ8hR9mlH2NM6kmZRFBaVcpd8+7io00fEdQgBZ0KuH/o/YzsOZK2aW2THZ4xxiRNyiSCVr5WBDTAuLxxjDpmFN/J/k6yQzLGmENCyiQCEeF3p/8u2WEYY8whx74Ga4wxKc4SgTHGpDhLBMYYk+IsERhjTIqzRGCMMSnOEoExxqQ4SwTGGJPiLBEYY0yKs0RgjDEpzhKBMcakOEsExhiT4iwRGGNMirNEYIwxKc4SgTHGpDhLBMYYk+IsERhjTIqzRGCMMSnOEoExxqQ4SwTGGJPiEpoIRGSkiHwuImtEZEKM8T1EZI6IfCoiy0Tk7ETGY4wxpq6EJQIR8QJPAWcBfYBLRaRPrWr3AdNUtRC4BHg6UfEYY4yJLZFnBIOBNaq6VlUrganA6Fp1FGjr9mcDmxMYjzHGmBgSmQi6Ahuihje6ZdEeBK4QkY3ATOCW+mYmIteLyEIRWbh9+/amjtUYY1JWsm8WXwq8qKrdgLOBP4pIzJhUdYqqDlTVgZ06dWrWII0x5nCWyESwCegeNdzNLYt2DTANQFXnAxlAxwTGZIwxppZEJoIFwLEi0lNE0nBuBr9dq87XwAgAEemNkwjsuo8xxjSjhCUCVQ0ANwOzgFU4TwetEJGHRGSUW+1/getEZCnwGjBOVTVRMRljjKnLl8iZq+pMnJvA0WUPRPWvBIYlMgZjjDENS/bNYmOMMUlmicAYY1KcJQJjjElxlgiMMSbFWSIwxpgUZ4nAGGNSnCUCY4xJcZYIjDEmxVkiMMaYFGeJwBhjUpwlAmOMSXGWCIwxJsVZIjDGmBRnicAYY1KcJQJjjElxlgiMMSbFWSIwxpgUZ4nAGGNSnCUCY4xJcZYIjDEmxTU6EYjIr0UkL5HBGGOMaX7xnBGsAqaIyCcicoOIZCcqKGOMMc3H19iKqvoc8JyIHA+MB5aJyL+AZ1V1TqICNMY0rKqqio0bN1JeXp7sUMwhICMjg27duuH3+xs9TaMTAYCIeIFebrcDWArcLiL/o6qXxDMvY0zT2LhxI23atCE3NxcRSXY4JolUlZ07d7Jx40Z69uzZ6OniuUfwG2A1cDbwC1UdoKq/VNVzgcK4IzbGNIny8nI6dOhgScAgInTo0CHus8N4zgiWAfep6r4Y4wbHtVRjTJOyJGDCDuR3IZ6bxbuJShwi0k5EzgNQ1eK4l2yMOSzs3r2bp59++oCmPfvss9m9e3eDdR544AFmz559QPOvLTc3lx07djTJvA4n8SSCidE7fFXdDUxs+pCMMS1JQ4kgEAg0OO3MmTNp165dg3UeeughTj/99AOOz+xfPIkgVt24bjYbYw4/EyZM4Msvv6R///7ceeedzJ07l5NOOolRo0bRp08fAM477zwGDBhAXl4eU6ZMiUwbPkJfv349vXv35rrrriMvL48zzzyTsrIyAMaNG8f06dMj9SdOnEhRURF9+/Zl9erVAGzfvp0zzjiDvLw8rr32Wo4++uj9HvlPmjSJ/Px88vPzmTx5MgD79u3jnHPOoV+/fuTn5/P6669H1rFPnz4UFBRwxx13NO0GPATEsyNfKCKTgKfc4ZuARU0fkjHmQP30rytYuXlPk86zT5e2TDy3/u+SPvLIIyxfvpwlS5YAMHfuXBYvXszy5csjT648//zztG/fnrKyMgYNGsSFF15Ihw4daszniy++4LXXXuPZZ5/l4osv5o033uCKK66os7yOHTuyePFinn76aR577DGee+45fvrTn3Laaadxzz338N577/GHP/yhwXVatGgRL7zwAp988gmqypAhQzjllFNYu3YtXbp04W9/+xsAxcXF7Ny5kxkzZrB69WpEZL+XslqieM4IbgEqgdfdrgInGdRLREaKyOciskZEJtRT52IRWSkiK0Tk1TjiMcYcogYPHlzj8cXHH3+cfv36MXToUDZs2MAXX3xRZ5qePXvSv39/AAYMGMD69etjzvuCCy6oU+ejjz7ikkucJ9hHjhxJTk5Og/F99NFHnH/++WRlZdG6dWsuuOACPvzwQ/r27cv777/P3XffzYcffkh2djbZ2dlkZGRwzTXX8Je//IXMzMx4N8chL54vlO0DYu7MY3G/c/AUcAawEVggIm+r6sqoOscC9wDDVHWXiHRudOTGmDoaOnJvTllZWZH+uXPnMnv2bObPn09mZibDhw+P+Xhjenp6pN/r9UYuDdVXz+v17vceRLyOO+44Fi9ezMyZM7nvvvsYMWIEDzzwAP/5z3/4xz/+wfTp03nyySf54IMPmnS5yRbP9wg6icijIjJTRD4Idw1MMhhYo6prVbUSmAqMrlXnOuApVd0FoKrb4l0BY0xytWnThpKSknrHFxcXk5OTQ2ZmJqtXr+bjjz9u8hiGDRvGtGnTAPj73//Orl27Gqx/0kkn8eabb1JaWsq+ffuYMWMGJ510Eps3byYzM5MrrriCO++8k8WLF7N3716Ki4s5++yz+c1vfsPSpUubPP5ki+cewSs4l4R+ANwAXA1sb6B+V2BD1PBGYEitOscBuE1VeIEHVfW9WDMTkeuB6wF69OgRR9jGmETq0KEDw4YNIz8/n7POOotzzjmnxviRI0fyu9/9jt69e3P88cczdOjQJo9h4sSJXHrppfzxj3/khBNO4Mgjj6RNmzb11i8qKmLcuHEMHux8Beraa6+lsLCQWbNmceedd+LxePD7/TzzzDOUlJQwevRoysvLUVUmTZrU5PEnm6hq4yqKLFLVASKyTFUL3LIFqjqonvoXASNV9Vp3+EpgiKreHFXnHaAKuBjoBswD+rqPptZr4MCBunDhwkbFbczhbtWqVfTu3TvZYSRVRUUFXq8Xn8/H/PnzufHGGyM3r1NRrN8Jdx8+MFb9eM4IqtzPb0TkHGAz0L6B+puA7lHD3dyyaBuBT1S1ClgnIv8FjgUWxBGXMSbFff3111x88cWEQiHS0tJ49tlnkx1SixJPIvi52/T0/wJPAG2BnzRQfwFwrIj0xEkAlwCX1arzJnAp8IKIdMS5VLQ2jpiMMYZjjz2WTz/9NNlhtFiNSgTuE0DHquo7QDFw6v6mUdWAiNwMzMK5/v+8qq4QkYeAhar6tjvuTBFZCQSBO1V15wGuizHGmAPQqESgqkERuRT4TTwzV9WZwMxaZQ9E9Stwu9sZY4xJgnguDf1LRJ7EeXIo0gKpqi5u8qiMMcY0m3gSQX/386GoMgVOa7pwjDHGNLdGf6FMVU+N0VkSMMbErXXr1gBs3ryZiy66KGad4cOHs7/HxCdPnkxpaWlkuDHNWjfGgw8+yGOPPXbQ82kpGn1GICIPxCpX1YdilRtjzP506dIl0rLogZg8eTJXXHFFpP2fmTNn7mcKE0s8jc7ti+qCwFlAbgJiMsa0IBMmTOCpp56KDIePpvfu3cuIESMiTUa/9dZbdaZdv349+fn5AJSVlXHJJZfQu3dvzj///BptDd14440MHDiQvLw8Jk50XoPy+OOPs3nzZk499VROPdV5kDH6xTOxmpluqLnr+ixZsoShQ4dSUFDA+eefH2m+4vHHH480TR1u8O6f//wn/fv3p3///hQWFjbY9MahJJ5G534dPSwij+E8/mmMOVS8OwG2fNa08zyyL5z1SL2jx44dy2233cZNNzmNEU+bNo1Zs2aRkZHBjBkzaNu2LTt27GDo0KGMGjWq3lcpPvPMM2RmZrJq1SqWLVtGUVFRZNzDDz9M+/btCQaDjBgxgmXLlnHrrbcyadIk5syZQ8eOHWvMq75mpnNychrd3HXYVVddxRNPPMEpp5zCAw88wE9/+lMmT57MI488wrp160hPT49cjnrsscd46qmnGDZsGHv37iUjI6PRmzmZ4jkjqC0T59vCxpgUVlhYyLZt29i8eTNLly4lJyeH7t27o6rce++9FBQUcPrpp7Np0ya2bt1a73zmzZsX2SEXFBRQUFAQGTdt2jSKioooLCxkxYoVrFy5sr7ZAPU3Mw2Nb+4anAbzdu/ezSmnnALA1Vdfzbx58yIxXn755fzpT3/C53OOqYcNG8btt9/O448/zu7duyPlh7p47hF8hvOUEDhfEOtEzSeIjDHJ1sCReyKNGTOG6dOns2XLFsaOHQvAK6+8wvbt21m0aBF+v5/c3NyYzU/vz7p163jsscdYsGABOTk5jBs37oDmE9bY5q73529/+xvz5s3jr3/9Kw8//DCfffYZEyZM4JxzzmHmzJkMGzaMWbNm0atXrwOOtbnEc0bwA+BctzsT6KKqTyYkKmNMizJ27FimTp3K9OnTGTNmDOAcTXfu3Bm/38+cOXP46quvGpzHySefzKuvOu+mWr58OcuWLQNgz549ZGVlkZ2dzdatW3n33Xcj09TXBHZ9zUzHKzs7m5ycnMjZxB//+EdOOeUUQqEQGzZs4NRTT+WXv/wlxcXF7N27ly+//JK+ffty9913M2jQoMirNA918Zy3HAWsUNUSABFpIyJ9VPWTxIRmjGkp8vLyKCkpoWvXrhx11FEAXH755Zx77rn07duXgQMH7vfI+MYbb2T8+PH07t2b3r17M2DAAAD69etHYWEhvXr1onv37gwbNiwyzfXXX8/IkSPp0qULc+bMiZTX18x0Q5eB6vPSSy9xww03UFpayne+8x1eeOEFgsEgV1xxBcXFxagqt956K+3ateP+++9nzpw5eDwe8vLyOOuss+JeXjLE0wz1p0CR2ywEIuLBaTOoqOEpm541Q21MNWuG2tQWbzPU8VwaEo3KGqoaIr4zCmOMMYegeBLBWhG5VUT8bvdjrMloY4xp8eJJBDcAJ+K8WyD82snrExGUMcaY5hPPF8q24bxcxhhjzGGk0WcEIvKSiLSLGs4RkecTE5YxxpjmEs+loYLol8qr6i6gsOlDMsYY05ziSQQeEckJD4hIe+ypIWOMafHi2ZH/GpgvIn8GBLgI+EVCojLGmFoCgUCLabunpYnnxTQvAxcAW4EtwAVumTEmxZ133nkMGDCAvLw8pkyZAsB7771HUVER/fr1Y8SIEQDs3buX8ePH07dvXwoKCnjjjTeA6hfVAEyfPp1x48YBMG7cOG644QaGDBnCXXfdxX/+8x9OOOEECgsLOfHEE/n8888BCAaD3HHHHeTn51NQUMATTzzBBx98wHnnnReZ7/vvv8/555/fHJujxYkrvarqSmCliHwXuExE/qyqeYkJzRgTr1/+55es/rZp27fp1b4Xdw++u8E6zz//PO3bt6esrIxBgwYxevRorrvuOubNm0fPnj359ttvAfjZz35GdnY2n33mNJUdbtu/IRs3buTf//43Xq+XPXv28OGHH+Lz+Zg9ezb33nsvb7zxBlOmTGH9+vUsWbIEn8/Ht99+S05ODj/60Y/Yvn07nTp14oUXXuCHP/zhwW+Qw1A8rY92AcYClwF9gf/DHic1xuC8pGXGjBkAbNiwgSlTpnDyySfTs2dPANq3bw/A7NmzmTp1amS6nJycujOrZcyYMXi9XsBpyO7qq6/miy++QESoqqqKzPeGG26IXDoKL+/KK6/kT3/6E+PHj2f+/Pm8/LJdxIhlv4lARK4HLgW6AtOAa4C3VPWnCY7NGBOn/R25J8LcuXOZPXs28+fPJzMzk+HDh9O/f/+4Wt6MfllN7Sams7KyIv33338/p556KjNmzGD9+vUMHz68wfmOHz+ec889l4yMDMaMGWP3GOrRmHsET7r1LlPV+1R1GdXvJTDGpLji4mJycnLIzMxk9erVfPzxx5SXlzNv3jzWrVsHELk0dMYZZ9R4rWX40tARRxzBqlWrCIVCkTOL+pbVtWtXAF588cVI+RlnnMHvf/97AoFAjeV16dKFLl268POf/5zx48c33UofZhqTCI4CXgN+LSKfi8jPAH9iwzLGtBQjR44kEAjQu3dvJkyYwNChQ+nUqRNTpkzhggsuoF+/fpGX1dx3333s2rWL/Px8+vXrF2k6+pFHHuEHP/gBJ554YqQZ61juuusu7rnnHgoLCyM7fXCame7RowcFBQX069cv8l4DcJrD7t69u7XQ2oBGN0MNICLdcO4TXApkATNU9d4ExVYva4bamGrWDHXDbr75ZgoLC7nmmmuSHUqzSWQz1KjqRlX9tTuz0cCBvy/OGGMSbMCAASxbtqzBl9Obg/hmsKr+F3tnsTHmELZo0aJkh9AiNOapoXXUvTkcvsWvbr8Ck1X18aYNzxhjTKLtNxGoas/mCMQYY0xyxNMMdZb7nmJE5DgRGSUi9vSQMca0cPHcLJ4HZIhIV+DvwJXAiw1NICIj3UdO14jIhAbqXSgiKiIx72gbY4xJnHhfXl+K0/Dc06o6Bqi3nSER8QJPAWcBfYBLRaRPjHptgB8Dn8QTuDGm5Qo3Mrd582YuuuiimHWGDx/O/h4Tnzx5MqWlpZHhs88+m927dzcwhYklrkQgIicAlwN/c8u8DdQfDKxR1bWqWglMxXnktLafAb/EHkU1JuV06dKF6dOnH/D0tRPBzJkzadeuXQNTHFpUlVAolOww4koEtwH34HyJbIWIfAeY00D9rsCGqOGNblmEiBQB3VX1b+yHiFwvIgtFZOH27dvjCNsYk0gTJkyo0WzEgw8+yGOPPcbevXsZMWIERUVF9O3bl7feeqvOtOvXryc/Px+AsrIyLrnkEnr37s35559PWVlZpN6NN97IwIEDycvLY+39CxQAABj6SURBVOLEiYDT0N3mzZs59dRTOfXUUwHIzc1lx44dAEyaNIn8/Hzy8/OZPHlyZHm9e/fmuuuuIy8vjzPPPLPGcsL++te/MmTIEAoLCzn99NPZunUrUH8z2rGa3A5vh7D8/HzWr1/P+vXrOf7447nqqqvIz89nw4YNMdcPYMGCBZx44on069ePwYMHU1JSwsknn8ySJUsidb73ve+xdOnSRv+8Yonn5fX/BP4J4N403qGqtx7ogt15TALGNXL5U4Ap4Hyz+ECXa8zhbMsvfkHFqqZthjq9dy+OvLf+BgTGjh3Lbbfdxk033QTAtGnTmDVrFhkZGcyYMYO2bduyY8cOhg4dyqhRo2o0MBftmWeeITMzk1WrVrFs2TKKiooi4x5++GHat29PMBhkxIgRLFu2jFtvvZVJkyYxZ84cOnbsWGNeixYt4oUXXuCTTz5BVRkyZAinnHIKOTk5fPHFF7z22ms8++yzXHzxxbzxxht1vnD2ve99j48//hgR4bnnnuNXv/oVv/71r2M2o719+/aYTW435IsvvuCll15i6NCh9a5fr169GDt2LK+//jqDBg1iz549tGrVimuuuYYXX3yRyZMn89///pfy8nL69eu332U2JJ6nhl4VkbYikgUsx3kvwZ0NTLIJ6B413M0tC2sD5ANzRWQ9MBR4224YG9OyFBYWsm3bNjZv3szSpUvJycmhe/fuqCr33nsvBQUFnH766WzatClyZB3LvHnzIjvkgoICCgoKIuOmTZtGUVERhYWFrFixgpUrVzYY00cffcT5559PVlYWrVu35oILLuDDDz8EoGfPnvTv3x9wvnm8fv36OtNv3LiR73//+/Tt25dHH32UFStWAE5z1+GEB04z2h9//HHMJrcbcvTRR0eSQH3r9/nnn3PUUUcxaNAgANq2bYvP52PMmDG88847VFVV8fzzz0de4nMw4vlmcR9V3SMilwPvAhOARcCj9dRfABwrIj1xEsAlOO8yAEBVi4FIGheRucAdqmqNCBlzgBo6ck+kMWPGMH36dLZs2RJpYO6VV15h+/btLFq0CL/fT25ubp0mphtj3bp1PPbYYyxYsICcnBzGjRt3QPMJS09Pj/R7vd6Yl4ZuueUWbr/9dkaNGsXcuXN58MEH416Oz+ercf0/OuboprXjXb/MzEzOOOMM3nrrLaZNm9Yk356O5x6B3/3ewHnA26paRQPNUatqALgZmAWsAqa59xYeEpFRBxO0MebQMnbsWKZOncr06dMZM2YM4DQZ3blzZ/x+P3PmzOGrr75qcB4nn3xypNXQ5cuXs2zZMgD27NlDVlYW2dnZbN26lXfffTcyTZs2bSgpKakzr5NOOok333yT0tJS9u3bx4wZMzjppJMavT7RzV2/9NJLkfJYzWgPHTo0ZpPbubm5LF68GIDFixdHxtdW3/odf/zxfPPNNyxYsACAkpKSSIur1157LbfeeiuDBg1q1Mt99ieeM4LfA+uBpcA8ETka2NPQBKo6E5hZq+yBeuoOjyMWY8whJC8vj5KSErp27RppRvryyy/n3HPPpW/fvgwcOJBevXo1OI8bb7yR8ePH07t3b3r37s2AAQMA6NevH4WFhfTq1Yvu3bszbNiwyDTXX389I0eOpEuXLpEmrQGKiooYN24cgwcPBpwdZ2FhYczLQLE8+OCDjBkzhpycHE477bTITvy+++7jpptuIj8/H6/Xy8SJE7ngggsiTW6HQiE6d+7M+++/z4UXXsjLL79MXl4eQ4YM4bjjjou5rPrWLy0tjddff51bbrmFsrIyWrVqxezZs2ndujUDBgygbdu2TfaOhbiaoa4zsYjPPfJvVtYMtTHVrBnq1LN582aGDx/O6tWr8XjqXthJWDPUIpItIpPCj3CKyK9x3klgjDGmmbz88ssMGTKEhx9+OGYSOBDxzOV5oAS42O32AC80SRTGGGMa5aqrrmLDhg2RezFNIZ57BN9V1Qujhn8qIkvqrW2MMaZFiOeMoExEvhceEJFhQN3nrowxze5g7vWZw8uB/C7Ec0ZwA/CyiGS7w7uAq+NeojGmSWVkZLBz5046dOhQ77d2TWpQVXbu3ElGRkZc08XTxMRSoJ+ItHWH94jIbcCyuJZojGlS3bp1Y+PGjVgbXAacA4Nu3brFNU3c7yxW1ejvDtwOTI53HsaYpuP3+yPNGxhzIA722SM7DzXGmBbuYBOB3aEyxpgWbr+XhkSkhNg7fAFaNXlExhhjmtV+E4GqtmmOQIwxxiRH03w/2RhjTItlicAYY1KcJQJjjElxKZUIgiG1r+IbY0wtKZMIikurGPv7+UxdsCHZoRhjzCElZRJBmwwfrdK8PPj2ClZ90+CL1YwxJqWkTCLweIRJF/enbSs/N726mH0Vzf5iNWOMOSSlTCIA6NQmnd9e0p/1O/Zx/1vLkx2OMcYcElIqEQCc+N2O3DriWP6yeBN/Xmj3C4wxJuUSAcAtpx3LCd/pwANvreCLrSXJDscYY5IqJROB1yP89pL+ZKV7uenVxZRVBpMdkjHGJE1KJgKAzm0z+M3Y/nyxbS8Pvr0i2eEYY0zSpGwiADjp2E78aPh3eX3hBt78dFOywzHGmKRI6UQA8JPTj2NQbg73zviMtdv3JjscY4xpdimfCHxeD49fWki6z8NNr35KeZXdLzDGpJaUTwQAR2W3YtLF/Vn1zR5+/reVyQ7HGGOalSUC16m9OvM/J3+HP338Ne8s25zscIwxptlYIohyx/ePp7BHOya88Rlf7dyX7HCMMaZZWCKI4vd6eOLSQrwe4eZXP6UiYPcLjDGHv4QmAhEZKSKfi8gaEZkQY/ztIrJSRJaJyD9E5OhExtMY3XIyefSiAj7bVMz/zVyd7HCMMSbhEpYIRMQLPAWcBfQBLhWRPrWqfQoMVNUCYDrwq0TFE48z845k/LBcXvz3et5bviXZ4RhjTEIl8oxgMLBGVdeqaiUwFRgdXUFV56hqqTv4MdAtgfHE5Z6zelPQLZu7pi9lw7el+5/AGGNaqEQmgq5AdPOeG92y+lwDvFvfSBG5XkQWisjC7du3N1GI9UvzeXjy0iJU4ZbXPqUyEEr4Mo0xJhkOiZvFInIFMBB4tL46qjpFVQeq6sBOnTo1S1w9OmTyy4sKWLJhN4/OsvsFxpjDUyITwSage9RwN7esBhE5Hfh/wChVrUhgPAfk7L5HceXQo3n2w3X8Y9XWZIdjjDFNLpGJYAFwrIj0FJE04BLg7egKIlII/B4nCWxLYCwH5f+d05s+R7Xlf/+8lM27y5IdjjHGNKmEJQJVDQA3A7OAVcA0VV0hIg+JyCi32qNAa+DPIrJERN6uZ3ZJleH38tTlRVQFQtz62qdUBe1+gTHm8CGqmuwY4jZw4EBduHBhsy/3rSWb+PHUJfxo+He5a2SvZl++McYcKBFZpKoDY407JG4WtxSj+3flkkHdeXrul/zzv4l/cskYY5qDJYI4TTw3j+OPaMPtry9h657yZIdjjDEHzRJBnFqleXnq8kJKK4Pc+tqnBEMt79KaMcZEs0RwAI7p3IafnZfPJ+u+5bf/+CLZ4RhjzEGxRHCALhrQjQuLuvHEB1/wrzU7kh2OMcYcMEsEB+Fn5+XxnY5Z/HjqEraXHHLfhTPGmEaxRHAQMtN8PHV5ESXlVfzk9SV2v8AY0yJZIjhIvY5sy09H5fHRmh08PWdNssMxxpi4WSJoAmMHdWdUvy78ZvZ/+WTtzmSHY4wxcbFE0AREhF9c0JejO2Rx69RP2bnX7hcYY1oOSwRNpHW6jycvK2RXaRW3T1tKyO4XGGNaCEsETSivSzb3n9Obf/53O1M+XJvscIwxplEsETSxK4Yezdl9j+TRWZ+z6Ktvkx2OMcbslyWCJiYiPHJhAV3aZXDLq5+ya19lskMyxpgGWSJIgLYZfp66rIjteyu4c/pSWmJT38aY1GGJIEEKurXjnrN6M3vVNv7w0bpkh2OMMfWyRJBA44flckafI/jle6tZsmF3ssMxxpiYLBEkkIjw6EUFdG6Twc2vLqa4rCrZIRljTB2WCBKsXWYaT1xWyJbicu6evszuFxhjDjmplQg2LYIty+HbdbB3G1Tug2bYMRf1yOHO7x/Peyu28MePv0r48owxJh6+ZAfQrF78AVSV1ioU8GdCWlbdzp8Jaa3rGRejLLrc3wpEIku57qTv8PHanfz8nVUU9cghv2t28667McbUI7USwdg/OWcBlfugcq+TFCLD+2qOqyiBki1Of+U+qCyFQFkcC5MaCcLjz2KKrxUL0yrZ8XwrNh3dnYx2R9A650jS23aCrA6Q2QEyOzqfaZkJ2wzGGBMttRLBMSMObvpQ0EkKkQQSlSTC/VVR/bU6f9U++naoYMv2r0n7cgXtKMErsS9NBb2tCLVqj6d1R7xZHSHLTRDhLqtjddLI6ggZ7cCTWlf6jDnshILOQWidbo/zKQJFVzX5YlMrERwsjxcy2jrdAWoNdNhXydod+5i/ax/bt2+leMcWSndvpbJ4G8F9O8kOFZMTKKF95R7aF5fQybuejp7ltNNiMrQ89ozFA63aRyWJ9jUTRZ0k0sG5fGXMoUgVAuXOQZR4nL89j6+6E0+NS69JF6h0ryS4O+zyPXV34rF27LW7qn0NLyezoyWCw0VOVhoDstIYcHQO0K3GOFVlx95KNu0uY+OuUj7fVcYHu5z+jbvK2LG7mIzK3bSXEtrLHnIooYt/Hz3Sy+ji3UfnwF5yivfQ+ttVtKrahbdiF6Kh2IH4WoE3reYfmddX8w+udldnvBc8/prDXn9848XjHAlpyPkMBUCDUWUBt98ti+5Xt34oFDU+0EDd2v3huu428niidjbh7eKpNeyt3ma16zS2rKF5ef3gywBfuvsZ3V/r0+s/xHaIFTV3bOHLrI0uc3emlXudn01DaiSGqO1Y43ev9rB3P+PrmSfUirvWjjxQzwFaNPFAehtIb+t+tnEO2HKOrlteo6tVngCWCA4xIkKnNul0apNO/+7t6oxXVXaVVrFxVymbdpWx0U0SH+wO95ext6L6D0gIcWRaOb3bVnJs6wqOziina/o+jvDtI0f20cobIk2CpHkUjwaqd47BquodcCgAoSqnP1gFVWVunYA7rr4uGDWfg/wOhXhq7jzFW2un6o36w/XW3PFK7Z2tF8Rfa15eZzk1koubYAKVoGVR2yM6gbh1wvVrl4WH60vGB0s8DSeKej/3U8eb7twTq9hba0e9J0ZZVNfYn3NaG0hvXb1zS2sNWZ1q7vDCD2pA3d+rOsOxymIMR34+tX6XY04T9XNFa+6Y2xwFHY9reIddu8yfeWgl7SiWCFoYEaF9Vhrts9Io6BY7UewpC7DBPYMIn1ls3FXGvF1lbNpSyp7y2EdabTJ8tMv0k93KT7tWaWS38pMdGfZHxmW748LDmWlepDG/4KGQm1Bq/+EF69mRh3feh9hlgAOhGntnFHMHVekcYQYq9vPZmDruEXr0cFWZO1wWX4LyZ9bcSae3gXY96pbV3pmnt6250/dn2f2sQ4wlgsOMiDg778zseh9RLS6rYtOuMr4pLmN3aRXFZVXsLqtiT1kVu0srI8Obi8sodscHGnjRjt8rboJwk0ZmWq3hmp9OIskgu5WfNF+K7BBEnMtq3kPsTy4YqD+5+FtFHZm3PvRiN03GfrIpKLyD7tOlcTe9VZV9lUEnQbiJojgqgTjlbiIpq2RbSTn/3VpCcVkVJfWcfYRlpnnJbuUnK91HVprX+Uz30TrdR1a6l6w0X1SZOz5S5nXrOfXTfZ7GnZmYal4feFs7R+wmZVkiMPslIrR2d7Zd28X3pFEgGKKkPBBJGOFkssdNHuGyfZUB9lUE2VcR4Nt9pTWGKwKNu3zh9QiZadXJITq5RBJLVCKJJBa3LDPNS5rPQ5rXg9/9TPN5SHf7PR5LMubwlNBEICIjgd8CXuA5VX2k1vh04GVgALATGKuq6xMZk2lePq+HnKw0crLSDngeVcEQpRVB9lYGKK0IsLfCSRLOZ4DSygB73aRRXVY9PpxYSt1pGptY6qyLR5xE4fPg9zrJId0dTotKHH5vdVm6t+b4Ogkm1nReDz6v4PN48HsFn9eDzyP4I+VOmd/99HkFv6d6nJ0VmXglLBGIiBd4CjgD2AgsEJG3VXVlVLVrgF2qeoyIXAL8EhibqJhMy+T3esjO9JCd6W+S+dWXWMqqglQGQlQG3c9AiMqgRvVHl4eoDKj7GYwqC7GvIuBOF4yURU9XFUxs+1ZOoohKDjGShtcj9SSZmsnHI4LX45xteT2CVwSP++n1up9uPZ/HHeepLvd6qutXj8et76nu94aXFXtar6d6vEeoWR6JixplHqlbbkkytkSeEQwG1qjqWgARmQqMBqITwWjgQbd/OvCkiIhaE50mgZo6scRLVWsmiKj+ikCIQEgJuAkjEAoRCCpVQae8KugMB0Lu+Ei5O00ouixEMGpczOndz9LKQI35BELOuFAIgiElqOp8hpSQOxyI6m8pf7Ei1ExmHnHKaic5T3UCCScVdddTcX6GCoTCZe76R4apVTeqP6TV06PuNBCZLqROeXge0eM7tk5n4X2nN/l2SWQi6ApsiBreCAypr46qBkSkGOgA7EhgXMYklYiQ7vOS7vMmO5QmE4pKFqGopBFOIqEQ1Yml1vhw/UBUf8gddqaNrufsGKOXES6vXTe6fu1lheuGlBrl1XWpUVdVEREEIp8eqe5HnDMbZzwIToIRcT/dck/UPMAdrjWeqOk90fMCstITs8tuMTeLReR64HqAHj16JDkaY0w0j0fwIPgPn9yWUhL5EPcmoHvUcDe3LGYdEfEB2Tg3jetQ1SmqOlBVB3bq1CkB4RpjTGpKZCJYABwrIj1FJA24BHi7Vp23gavd/ouAD+z+gDHGNK+EXRpyr/nfDMzCeXz0eVVdISIPAQtV9W3gD8AfRWQN8C1OsjDGGNOMEnqPQFVnAjNrlT0Q1V8OjElkDMYYYxqWIg29GGOMqY8lAmOMSXGWCIwxJsVZIjDGmBQnLfFpTRHZDnx1gJN3xL65HGbboibbHjXZ9qh2OGyLo1U15pewWmQiOBgislBVByY7jkOBbYuabHvUZNuj2uG+LezSkDHGpDhLBMYYk+JSMRFMSXYAhxDbFjXZ9qjJtke1w3pbpNw9AmOMMTWl4hmBMcaYKJYIjDEmxaVMIhCRkSLyuYisEZEJyY4nmUSku4jMEZGVIrJCRH6c7JiSTUS8IvKpiLyT7FiSTUTaich0EVktIqtE5IRkx5RMIvIT9+9kuYi8JiIZyY6pqaVEIhARL/AUcBbQB7hURPokN6qkCgD/q6p9gKHATSm+PQB+DKxKdhCHiN8C76lqL6AfKbxdRKQrcCswUFXzcZrUP+yay0+JRAAMBtao6lpVrQSmAqOTHFPSqOo3qrrY7S/B+UPvmtyokkdEugHnAM8lO5ZkE5Fs4GScd4WgqpWquju5USWdD2jlvkUxE9ic5HiaXKokgq7AhqjhjaTwji+aiOQChcAnyY0kqSYDdwGhZAdyCOgJbAdecC+VPSciWckOKllUdRPwGPA18A1QrKp/T25UTS9VEoGJQURaA28At6nqnmTHkwwi8gNgm6ouSnYshwgfUAQ8o6qFwD4gZe+piUgOztWDnkAXIEtErkhuVE0vVRLBJqB71HA3tyxliYgfJwm8oqp/SXY8STQMGCUi63EuGZ4mIn9KbkhJtRHYqKrhM8TpOIkhVZ0OrFPV7apaBfwFODHJMTW5VEkEC4BjRaSniKTh3Ox5O8kxJY2ICM414FWqOinZ8SSTqt6jqt1UNRfn9+IDVT3sjvgaS1W3ABtE5Hi3aASwMokhJdvXwFARyXT/bkZwGN48T+g7iw8VqhoQkZuBWTh3/Z9X1RVJDiuZhgFXAp+JyBK37F73HdPG3AK84h40rQXGJzmepFHVT0RkOrAY52m7TzkMm5uwJiaMMSbFpcqlIWOMMfWwRGCMMSnOEoExxqQ4SwTGGJPiLBEYY0yKs0RgTAwiEhSRJVFdk327VkRyRWR5U83PmIOVEt8jMOYAlKlq/2QHYUxzsDMCY+IgIutF5Fci8pmI/EdEjnHLc0XkAxFZJiL/EJEebvkRIjJDRJa6Xbh5Aq+IPOu2c/93EWmVtJUyKc8SgTGxtap1aWhs1LhiVe0LPInTcinAE8BLqloAvAI87pY/DvxTVfvhtNkT/kb7scBTqpoH7AYuTPD6GFMv+2axMTGIyF5VbR2jfD1wmqqudRvu26KqHURkB3CUqla55d+oakcR2Q50U9WKqHnkAu+r6rHu8N2AX1V/nvg1M6YuOyMwJn5aT388KqL6g9j9OpNElgiMid/YqM/5bv+/qX6F4eXAh27/P4AbIfJe5OzmCtKYxrKjEGNiaxXVMis47/ANP0KaIyLLcI7qL3XLbsF5q9edOG/4CrfY+WNgiohcg3PkfyPOm66MOWTYPQJj4uDeIxioqjuSHYsxTcUuDRljTIqzMwJjjElxdkZgjDEpzhKBMcakOEsExhiT4iwRGGNMirNEYIwxKe7/A5NNersseRQjAAAAAElFTkSuQmCC\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#  Đánh giá model với dữ liệu test set\n",
        "score = model.evaluate(X_test, Y_test, verbose=0)\n",
        "print(score)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ChLfmg1UK6fD",
        "outputId": "5cb4b452-53ab-40ae-cb69-cd8aabe7e5fa"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[0.04121463745832443, 0.9872000217437744]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Dự đoán ảnh\n",
        "plt.imshow(X_test[2000].reshape(28,28), cmap='gray')\n",
        "\n",
        "y_predict = model.predict(X_test[2000].reshape(1,28,28,1))\n",
        "print('Giá trị dự đoán: ', np.argmax(y_predict))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        },
        "id": "ZVfgcLqaLc5A",
        "outputId": "fb286a04-b294-445c-8a3f-197f26cb8c51"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1/1 [==============================] - 0s 20ms/step\n",
            "Giá trị dự đoán:  6\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPsAAAD4CAYAAAAq5pAIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAANpklEQVR4nO3dbayU9ZnH8d9PpS+ESnhQREqWtp431chDiG6yYtwQUCGKNbEBk41aEvqiGIhrFLsxJdk0ml27a3xDQgP2rOnakEgtokl7JHXdTbSKaBVlW12CAXLgRHnRUzWy4LUvzs3mqGf+c5hnzvX9JJOZua+5z1wZ/XE/zX/+jggBmPjO63YDADqDsANJEHYgCcIOJEHYgSQu6OSb2ebUP9BmEeGxlje1Zbd9o+0/2n7f9qZm/haA9nKj19ltny/pT5KWSToi6TVJayLi3cI6bNmBNmvHlv1qSe9HxMGIOCnpl5JWNfH3ALRRM2GfI+nwqOdHqmVfYHud7b229zbxXgCa1PYTdBGxVdJWid14oJua2bIflTR31PNvVMsA9KBmwv6apD7b37T9NUmrJe1qTVsAWq3h3fiIOGV7vaTfSDpf0vaIeKdlnQFoqYYvvTX0ZhyzA23Xli/VADh3EHYgCcIOJEHYgSQIO5AEYQeS6Oh4dmC0BQsWFOsbNmwo1u++++5WtjPhsWUHkiDsQBKEHUiCsANJEHYgCcIOJMGlNzRlypQpxfrGjRtr1j7++OPiujt37myoJ4yNLTuQBGEHkiDsQBKEHUiCsANJEHYgCcIOJMF1dhRdeeWVxfrDDz9crK9cubJm7cUXXyyuu3nz5mIdZ4ctO5AEYQeSIOxAEoQdSIKwA0kQdiAJwg4kwXV2FC1durRYL11Hl6R9+/bVrK1evbq47tDQULGOs9NU2G0fkjQs6bSkUxGxuBVNAWi9VmzZ/zYiPmzB3wHQRhyzA0k0G/aQ9Fvbr9teN9YLbK+zvdf23ibfC0ATmt2NvzYijtq+RNKA7f+OiJdGvyAitkraKkm2o8n3A9CgprbsEXG0uh+S9CtJV7eiKQCt13DYbU+2/fUzjyUtl7S/VY0BaC1HNLZnbftbGtmaSyOHA/8eET+psw678R02f/78Yv2aa64p1u+5555i/ZVXXinW77333pq14eHh4rpoTER4rOUNH7NHxEFJ5f+TAPQMLr0BSRB2IAnCDiRB2IEkCDuQBENcJ7hFixYV6+vXry/WH3300WK9v7//rHtCd7BlB5Ig7EAShB1IgrADSRB2IAnCDiRB2IEkuM4+wfX19RXrl19+ebF+3nlsDyYK/ksCSRB2IAnCDiRB2IEkCDuQBGEHkiDsQBIN/5R0Q2/GT0m3xYIFC2rWBgYGiuu++uqrxfptt91WrH/22WfFOjqv1k9Js2UHkiDsQBKEHUiCsANJEHYgCcIOJEHYgSQYz34OmDdvXrH+wAMP1KwNDg4W133ooYeKda6jTxx1t+y2t9sesr1/1LLptgdsv1fdT2tvmwCaNZ7d+J9LuvFLyzZJ2hMRfZL2VM8B9LC6YY+IlySd+NLiVZLOzPvTL+nWFvcFoMUaPWafFRFnDgaPSZpV64W210la1+D7AGiRpk/QRUSUBrhExFZJWyUGwgDd1Oilt+O2Z0tSdT/UupYAtEOjYd8l6c7q8Z2Sft2adgC0S93deNtPSbpe0kzbRyT9WNIjknbYXivpA0nfa2eT2d18883F+sqVK2vWSmPdJengwYMN9YRzT92wR8SaGqWlLe4FQBvxdVkgCcIOJEHYgSQIO5AEYQeSYIhrD7j00kuL9WXLlhXrb7zxRs1aty+tTZ06tWZt5syZxXXrDc/95JNPGuopK7bsQBKEHUiCsANJEHYgCcIOJEHYgSQIO5AEUzafA6677rpi/fHHH69ZW7JkSXHd4eHhYn3u3LnF+l133VWs33LLLTVrM2bMKK576NChYn3Xrl3F+pYtW2rWJvJPZDNlM5AcYQeSIOxAEoQdSIKwA0kQdiAJwg4kwXj2c8DFF19crF911VU1axs3biyu+8QTTxTrzzzzTLG+aNGiYr2Z73HMmTOnWJ80aVKxXrqWXroGP1GxZQeSIOxAEoQdSIKwA0kQdiAJwg4kQdiBJBjP3gPq/W58f39/sV76XfnnnnuuuO7y5cuL9U8//bRY37FjR7FeuhY+efLk4roPPvhgsX7BBeWviZTG+d9www3Fdc9lDY9nt73d9pDt/aOWbbZ91Pab1W1FK5sF0Hrj2Y3/uaQbx1j+rxGxoLo939q2ALRa3bBHxEuSTnSgFwBt1MwJuvW236p286fVepHtdbb32t7bxHsBaFKjYd8i6duSFkgalPTTWi+MiK0RsTgiFjf4XgBaoKGwR8TxiDgdEZ9L+pmkq1vbFoBWayjstmePevpdSftrvRZAb6g7nt32U5KulzTT9hFJP5Z0ve0FkkLSIUk/aGOPE96xY8eK9ZdffrlYL11nX7lyZXHdZ599tli//fbbi/WTJ08W6+100003FesDAwMd6uTcUDfsEbFmjMXb2tALgDbi67JAEoQdSIKwA0kQdiAJwg4kwRDXHnDJJZcU6wcOHCjWp02r+W3lupef7rjjjmL9o48+Ktbbad68ecV6vcuGK1bUHox5+PDhRlo6JzBlM5AcYQeSIOxAEoQdSIKwA0kQdiAJwg4kwZTNPeCKK64o1usNIy0NgZ05c2Zx3XZ/z6I03fSpU6eK6z7/fPl3THfv3l2sT+Rr6Y1gyw4kQdiBJAg7kARhB5Ig7EAShB1IgrADSXCdvQeUxl1L0pIlS4r1hQsX1qxt3769uO5jjz1WrN93333F+vz584v10pTOU6dOLa67Z8+eYv3+++8v1vFFbNmBJAg7kARhB5Ig7EAShB1IgrADSRB2IAl+N74H1Pt99OHh4WL99OnTNWuXXXZZcd0XXnihWL/wwguL9YsuuqhYL3nyySeL9bVr1xbr9cbDZ9Xw78bbnmv7d7bftf2O7Q3V8um2B2y/V93XnqkAQNeNZzf+lKS/j4jvSPprST+0/R1JmyTtiYg+SXuq5wB6VN2wR8RgROyrHg9LOiBpjqRVkvqrl/VLurVdTQJo3ll9N972PEkLJf1e0qyIGKxKxyTNqrHOOknrGm8RQCuM+2y87SmSnpa0MSL+PLoWI2f5xjz5FhFbI2JxRCxuqlMATRlX2G1P0kjQfxERO6vFx23PruqzJQ21p0UArVD30ptta+SY/EREbBy1/J8lfRQRj9jeJGl6RBTHHHLprffMmDGjWK83BLavr69YLw2x3bZtW3Hd0iVF1Fbr0tt4jtn/RtLfSXrb9pvVsh9JekTSDttrJX0g6XutaBRAe9QNe0T8l6Qx/6WQtLS17QBoF74uCyRB2IEkCDuQBGEHkiDsQBIMcQUmmIaHuAKYGAg7kARhB5Ig7EAShB1IgrADSRB2IAnCDiRB2IEkCDuQBGEHkiDsQBKEHUiCsANJEHYgCcIOJEHYgSQIO5AEYQeSIOxAEoQdSIKwA0kQdiCJumG3Pdf272y/a/sd2xuq5ZttH7X9ZnVb0f52ATSq7iQRtmdLmh0R+2x/XdLrkm7VyHzsf4mIR8f9ZkwSAbRdrUkixjM/+6CkwerxsO0Dkua0tj0A7XZWx+y250laKOn31aL1tt+yvd32tBrrrLO91/bepjoF0JRxz/Vme4qk/5D0k4jYaXuWpA8lhaR/1Miu/vfr/A1244E2q7UbP66w254kabek30TEv4xRnydpd0RcWefvEHagzRqe2NG2JW2TdGB00KsTd2d8V9L+ZpsE0D7jORt/raT/lPS2pM+rxT+StEbSAo3sxh+S9IPqZF7pb7FlB9qsqd34ViHsQPsxPzuQHGEHkiDsQBKEHUiCsANJEHYgCcIOJEHYgSQIO5AEYQeSIOxAEoQdSIKwA0kQdiCJuj842WIfSvpg1POZ1bJe1Ku99WpfEr01qpW9/VWtQkfHs3/lze29EbG4aw0U9GpvvdqXRG+N6lRv7MYDSRB2IIluh31rl9+/pFd769W+JHprVEd66+oxO4DO6faWHUCHEHYgia6E3faNtv9o+33bm7rRQy22D9l+u5qGuqvz01Vz6A3Z3j9q2XTbA7bfq+7HnGOvS731xDTehWnGu/rZdXv6844fs9s+X9KfJC2TdETSa5LWRMS7HW2kBtuHJC2OiK5/AcP2dZL+IunfzkytZfufJJ2IiEeqfyinRcQDPdLbZp3lNN5t6q3WNON3qYufXSunP29EN7bsV0t6PyIORsRJSb+UtKoLffS8iHhJ0okvLV4lqb963K+R/1k6rkZvPSEiBiNiX/V4WNKZaca7+tkV+uqIboR9jqTDo54fUW/N9x6Sfmv7ddvrut3MGGaNmmbrmKRZ3WxmDHWn8e6kL00z3jOfXSPTnzeLE3RfdW1ELJJ0k6QfVrurPSlGjsF66drpFknf1sgcgIOSftrNZqppxp+WtDEi/jy61s3Pboy+OvK5dSPsRyXNHfX8G9WynhARR6v7IUm/0shhRy85fmYG3ep+qMv9/L+IOB4RpyPic0k/Uxc/u2qa8acl/SIidlaLu/7ZjdVXpz63boT9NUl9tr9p+2uSVkva1YU+vsL25OrEiWxPlrRcvTcV9S5Jd1aP75T06y728gW9Mo13rWnG1eXPruvTn0dEx2+SVmjkjPz/SPqHbvRQo69vSfpDdXun271Jekoju3X/q5FzG2slzZC0R9J7kl6QNL2HentSI1N7v6WRYM3uUm/XamQX/S1Jb1a3Fd3+7Ap9deRz4+uyQBKcoAOSIOxAEoQdSIKwA0kQdiAJwg4kQdiBJP4PB85OIkRsQkAAAAAASUVORK5CYII=\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}