exam20210923_2022년2월21일 품.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "name": "exam20210923 (1).ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.6.8"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/kcw0331/Deeplearning/blob/main/exam20210923_2022%EB%85%842%EC%9B%9421%EC%9D%BC%20%ED%92%88.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZPi3ACwumNQV"
      },
      "source": [
        "# 석사 논문 자격시험 (파이썬을 이용한 딥러닝)\n",
        "\n",
        "- 이름:  김창우\n",
        "- 학번:  2020220071\n",
        "\n",
        "------------------------------------------\n",
        "\n",
        "colab을 사용할 경우 시험시작전에 꼭 Runtime tab을 클릭하고, Runtime - Change runtime type 을 클릭 후 Hardware accelerator 를 gpu로 설정해주세요.\n",
        "\n",
        "\n",
        "다음은 mnist 데이터를 활용한 CNN 모형에 관련된 문제입니다. 코드를 완성해주세요.\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 37
        },
        "id": "-6p6VRmTmAPZ",
        "outputId": "eecc9277-ea8b-4e2a-bce4-0d4133a1a3ff"
      },
      "source": [
        "import tensorflow as tf\n",
        "tf.keras.__version__"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            },
            "text/plain": [
              "'2.8.0'"
            ]
          },
          "metadata": {},
          "execution_count": 1
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "APqepHCVmnbd"
      },
      "source": [
        "데이터 읽어오기."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "g2lNQsZzmLsE",
        "outputId": "4418c6e1-b14e-4cd4-d6b6-84e5837c002f"
      },
      "source": [
        "from tensorflow.keras.datasets import mnist\n",
        "from tensorflow.keras.utils import to_categorical\n",
        "\n",
        "(x_train, y_train), (x_test, y_test) = mnist.load_data()\n",
        "x_train, x_test = x_train / 255.0, x_test / 255.0\n",
        "\n",
        "# Add a channels dimension\n",
        "x_train = x_train[..., tf.newaxis].astype(\"float32\")\n",
        "x_test = x_test[..., tf.newaxis].astype(\"float32\")\n",
        "\n",
        "train_labels = to_categorical(y_train)\n",
        "test_labels = to_categorical(y_test)"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz\n",
            "11493376/11490434 [==============================] - 0s 0us/step\n",
            "11501568/11490434 [==============================] - 0s 0us/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x_train.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qMcW52ykXazn",
        "outputId": "8cac3bae-f9d9-4084-f5af-11c2a420eb8d"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(60000, 28, 28, 1)"
            ]
          },
          "metadata": {},
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YmECe4jhmqgW"
      },
      "source": [
        "(1) [20점] 아래의 그림과 같은 구조를 가지는 CNN 모형을 작성해 봅시다. \n",
        "\n",
        "![image.png](data:image/png;base64,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)\n",
        "\n",
        "- Convolution layer 의 필터 갯수는 각각 16, 32, 32개로 설정하세요\n",
        "- 모든 Convolution layer 에서 커널 크기는 모두 (3,3) 으로 설정하고, zero padding을 적용하고. stride 는 1칸씩 진행되도록 합니다.\n",
        "- 모든 Convolution layer 에서 activation 은 'relu'를 사용합니다.\n",
        "- Maxpooling 은 가로 세로 2칸씩 적용하도록 하고, stride 역시 2칸씩 진행합니다.\n",
        "- Flatten layer 이후의 Dense layer 의 필터 갯수는 50으로 하고, activation은 'relu'함수를 사용합니다. \n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Gnq4bd9AmLkk"
      },
      "source": [
        "### Define your model here\n",
        "from keras import models\n",
        "from keras import layers\n",
        "\n",
        "model = models.Sequential()\n",
        "model.add(layers.Conv2D(16,(3,3), strides = (1,1), activation = 'relu', padding = 'valid' ,input_shape = (28, 28, 1)))\n",
        "model.add(layers.MaxPool2D(pool_size=(2,2), strides = (2,2)))\n",
        "model.add(layers.Conv2D(32,(3,3), strides = (1,1), activation = 'relu', padding = 'valid'))\n",
        "model.add(layers.MaxPool2D(pool_size=(2,2), strides = (2,2)))\n",
        "model.add(layers.Conv2D(32,(3,3), strides = (1,1), activation = 'relu', padding = 'valid'))\n",
        "model.add(layers.Flatten())\n",
        "model.add(layers.Dense(50, activation = 'relu'))\n",
        "model.add(layers.Dense(10, activation = 'softmax'))"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XDJo_v85nDUU"
      },
      "source": [
        "(2) [20점] 아래와 같이 모형을 컴파일하고 훈련을 수행해보세요.  \n",
        " - optimizer 로는 Adam optimizer 사용\n",
        " - loss는 categorical cross entropy를 사용\n",
        " - 모형 평가 metric 으로 accuracy를 사용\n",
        " - 훈련에 있어서 epoch=5, batch size 는 32를 사용 "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5bZ68LXLxMeJ",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9aebcb3d-4689-4765-d740-42d9d7ddc5e3"
      },
      "source": [
        "## compile your model with model.compile()\n",
        "model.compile(loss = 'categorical_crossentropy',\n",
        "              optimizer = 'Adam',\n",
        "              metrics = ['accuracy'])\n",
        "              \n",
        "## fit your model with model.fit()\n",
        "model.fit(x_train, train_labels, epochs = 5, batch_size = 32)"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/5\n",
            "1875/1875 [==============================] - 19s 5ms/step - loss: 0.1908 - accuracy: 0.9418\n",
            "Epoch 2/5\n",
            "1875/1875 [==============================] - 9s 5ms/step - loss: 0.0594 - accuracy: 0.9813\n",
            "Epoch 3/5\n",
            "1875/1875 [==============================] - 9s 5ms/step - loss: 0.0447 - accuracy: 0.9858\n",
            "Epoch 4/5\n",
            "1875/1875 [==============================] - 9s 5ms/step - loss: 0.0357 - accuracy: 0.9886\n",
            "Epoch 5/5\n",
            "1875/1875 [==============================] - 9s 5ms/step - loss: 0.0288 - accuracy: 0.9908\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7f3fb01ff910>"
            ]
          },
          "metadata": {},
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "b9O1gDaRzOuH"
      },
      "source": [
        "(3) [30점] Data Augmentation 을 적용하여 훈련을 시켜보세요. \n",
        "\n",
        "- Data Augmentation 으로 RandomFlip 을 가로, 세로방향을 적용하고, RandomRotation을 시계방향 반시계방향으로 30도 정도 회전을 적용합니다. \n",
        "- Tensorflow data api를 사용하면 조금 더 편한 것 같습니다. (안사용해도 괜찮습니다)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "O1N6AgBYzOuI",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "2c0640ad-83ab-436e-b4c2-8ee76e2a76fc"
      },
      "source": [
        "data_augmentation = tf.keras.Sequential([\n",
        "  layers.RandomFlip(\"horizontal_and_vertical\"),\n",
        "  layers.RandomRotation(0.3),\n",
        "])\n",
        "\n",
        "model1 = models.Sequential()\n",
        "model1.add(data_augmentation)\n",
        "model1.add(layers.Conv2D(16,(3,3), strides = (1,1), activation = 'relu', padding = 'valid' ,input_shape = (28, 28, 1)))\n",
        "model1.add(layers.MaxPool2D(pool_size=(2,2), strides = (2,2)))\n",
        "model1.add(layers.Conv2D(32,(3,3), strides = (1,1), activation = 'relu', padding = 'valid'))\n",
        "model1.add(layers.MaxPool2D(pool_size=(2,2), strides = (2,2)))\n",
        "model1.add(layers.Conv2D(32,(3,3), strides = (1,1), activation = 'relu', padding = 'valid'))\n",
        "model1.add(layers.Flatten())\n",
        "model1.add(layers.Dense(50, activation = 'relu'))\n",
        "model1.add(layers.Dense(10, activation = 'softmax'))\n",
        "\n",
        "model.compile(loss = 'categorical_crossentropy',\n",
        "              optimizer = 'Adam',\n",
        "              metrics = ['accuracy'])\n",
        "              \n",
        "model.fit(x_train, train_labels, epochs = 5, batch_size = 32)"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/5\n",
            "1875/1875 [==============================] - 10s 5ms/step - loss: 0.0251 - accuracy: 0.9920\n",
            "Epoch 2/5\n",
            "1875/1875 [==============================] - 9s 5ms/step - loss: 0.0202 - accuracy: 0.9935\n",
            "Epoch 3/5\n",
            "1875/1875 [==============================] - 9s 5ms/step - loss: 0.0175 - accuracy: 0.9943\n",
            "Epoch 4/5\n",
            "1875/1875 [==============================] - 10s 5ms/step - loss: 0.0156 - accuracy: 0.9951\n",
            "Epoch 5/5\n",
            "1875/1875 [==============================] - 9s 5ms/step - loss: 0.0139 - accuracy: 0.9955\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7f3fb029c490>"
            ]
          },
          "metadata": {},
          "execution_count": 15
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "z_MCHpg-zOuI"
      },
      "source": [
        "(4) [30점] 5이상의 숫자(5,6,7,8,9)와 5미만의 숫자(0,1,2,3,4)를 판별하는 모형을 만들고 싶습니다. 아래의 작업들을 수행하세요.\n",
        "\n",
        "- 5이상의 숫자를 0, 5미만의 숫자를 1로 하는 train_y_5, test_y_5 변수들을 만듭니다.\n",
        "- (2) 또는 (3)에서 적합시킨 pre-trained network 을 가져와서 transefer learning 을 적용해서 5이상의 숫자와 5미만 숫자를 판별하는 모형을 만들어보세요.   \n",
        "  - pre-trained network을 가져와서 마지막 layer 를 해당 문제에 맞게 수정해야 합니다.\n",
        "  - fine-tuning 을 적용합니다. 마지막 Dense(50) Lyaer 를 trainable 하게하고, 그 상위 CNN 레이어 들은 Freeze 시킨 후 트레이닝 해보세요. "
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "\n",
        "train_y_5 = []\n",
        "\n",
        "for i in y_train:\n",
        "  if i >= 5:\n",
        "    train_y_5.append('0')\n",
        "  else:\n",
        "    train_y_5.append('1')\n",
        "\n",
        "train_y_5 = np.array(train_y_5, dtype = np.uint8)\n",
        "\n",
        "test_y_5 = []\n",
        "\n",
        "for i in y_test:\n",
        "  if i >= 5:\n",
        "    test_y_5.append('0')\n",
        "  else:\n",
        "    test_y_5.append('1')\n",
        "\n",
        "test_y_5 = np.array(test_y_5, dtype = np.uint8)"
      ],
      "metadata": {
        "id": "47F2VJn8ehZ5"
      },
      "execution_count": 67,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "train_y_5_labels = to_categorical(train_y_5)\n",
        "test_y_5_labels = to_categorical(test_y_5)"
      ],
      "metadata": {
        "id": "zFR3S6png6uA"
      },
      "execution_count": 68,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "G4hHqDVphKeb",
        "outputId": "9b4f2b22-95d9-4739-8b77-59f8e25870fe"
      },
      "execution_count": 69,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_3\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " conv2d_9 (Conv2D)           (None, 26, 26, 16)        160       \n",
            "                                                                 \n",
            " max_pooling2d_6 (MaxPooling  (None, 13, 13, 16)       0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " conv2d_10 (Conv2D)          (None, 11, 11, 32)        4640      \n",
            "                                                                 \n",
            " max_pooling2d_7 (MaxPooling  (None, 5, 5, 32)         0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " conv2d_11 (Conv2D)          (None, 3, 3, 32)          9248      \n",
            "                                                                 \n",
            " flatten_3 (Flatten)         (None, 288)               0         \n",
            "                                                                 \n",
            " dense_6 (Dense)             (None, 50)                14450     \n",
            "                                                                 \n",
            " dense_7 (Dense)             (None, 10)                510       \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 29,008\n",
            "Trainable params: 0\n",
            "Non-trainable params: 29,008\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "a = model.layers[6]\n",
        "a.name"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 37
        },
        "id": "CO6BsYqfjHES",
        "outputId": "238e68bb-5de7-40d1-999b-bff079326c7d"
      },
      "execution_count": 70,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            },
            "text/plain": [
              "'dense_6'"
            ]
          },
          "metadata": {},
          "execution_count": 70
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model2 = models.Sequential()\n",
        "model2.add(model)\n",
        "model2.add(layers.Dense(2, activation = 'softmax'))"
      ],
      "metadata": {
        "id": "jofYuz13hucc"
      },
      "execution_count": 81,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "set_trainable = False\n",
        "for layer in model.layers:\n",
        "  if layer.name == model.layers[6].name:\n",
        "    set_trainable = True\n",
        "  if set_trainable:\n",
        "    layer.trainable = True\n",
        "  else:\n",
        "    layer.trainable = False"
      ],
      "metadata": {
        "id": "taHxT2j9j-uD"
      },
      "execution_count": 82,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model2.compile(loss = 'categorical_crossentropy',\n",
        "              optimizer = 'Adam',\n",
        "              metrics = ['accuracy'])\n",
        "              \n",
        "model2.fit(x_train, train_y_5_labels, epochs = 5, batch_size = 32)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1Ath-fqNj_zk",
        "outputId": "0c434234-3939-4608-8a74-975a4a316788"
      },
      "execution_count": 83,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/5\n",
            "1875/1875 [==============================] - 8s 4ms/step - loss: 0.2081 - accuracy: 0.9602\n",
            "Epoch 2/5\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.0408 - accuracy: 0.9955\n",
            "Epoch 3/5\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.0213 - accuracy: 0.9966\n",
            "Epoch 4/5\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.0143 - accuracy: 0.9973\n",
            "Epoch 5/5\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.0107 - accuracy: 0.9979\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7f3f43e57350>"
            ]
          },
          "metadata": {},
          "execution_count": 83
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "result = model2.evaluate(x_test, test_y_5_labels)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kApiSaaJpEBF",
        "outputId": "111efacf-aa38-4510-bbd4-bf3eb07fa08f"
      },
      "execution_count": 85,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "313/313 [==============================] - 1s 4ms/step - loss: 0.0250 - accuracy: 0.9942\n"
          ]
        }
      ]
    }
  ]
}