material/deep_learning/mnist_super_resolution.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "mnist_super_resolution.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "7u-zHeaOyjn2",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "c31a4307-3b52-4f02-d260-1917382c87ac"
      },
      "source": [
        "%tensorflow_version 1.x"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "TensorFlow 1.x selected.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jLKCaOxfGQg3",
        "colab_type": "text"
      },
      "source": [
        "modified from https://blog.naver.com/ckdgus1433/221443838135"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PhlLzK2qzOQU",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "from tensorflow.keras.datasets import mnist\n",
        "from tensorflow.keras.layers import Dense\n",
        "from tensorflow.keras.models import Sequential"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HHW1dtOvzOTu",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# MNIST 로딩 (라벨은 필요없기 때문에 버림)\n",
        "(x_train, _), (x_test, _) = mnist.load_data()"
      ],
      "execution_count": 34,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "E5B1_T2yzONm",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# 데이터 정규화 및 Reshape\n",
        "x_train = x_train.astype('float32') / 255.\n",
        "x_test = x_test.astype('float32') / 255."
      ],
      "execution_count": 35,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9I4G_rT-zecq",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 50
        },
        "outputId": "849450b5-864d-4917-9744-203e8f44ad3d"
      },
      "source": [
        "def shrink(data, rows, cols):\n",
        "    return data.reshape(data.shape[0], rows, int(data.shape[1]/rows), cols, int(data.shape[2]/cols)).sum(axis=2).sum(axis=3)\n",
        "\n",
        "reduced_x_train = shrink(x_train, 14, 14)\n",
        "reduced_x_test = shrink(x_test, 14, 14)\n",
        "print(reduced_x_train.shape)\n",
        "print(reduced_x_test.shape)\n"
      ],
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(60000, 14, 14)\n",
            "(10000, 14, 14)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5jFitrsf13K3",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 214
        },
        "outputId": "2d284781-6bb8-4187-f148-4b46a4075b5f"
      },
      "source": [
        "n = 10\n",
        "\n",
        "plt.figure(figsize=(20, 2))\n",
        "for i in range(1,n):\n",
        "    ax = plt.subplot(1, n, i)\n",
        "    plt.imshow(x_test[i])\n",
        "    plt.gray()\n",
        "    ax.get_xaxis().set_visible(False)\n",
        "    ax.get_yaxis().set_visible(False)\n",
        "plt.show()\n",
        "\n",
        "# 해상도 낮아진 이미지 확인\n",
        "plt.figure(figsize=(20, 2))\n",
        "for i in range(1,n):\n",
        "    ax = plt.subplot(1, n, i)\n",
        "    plt.imshow(reduced_x_test[i])\n",
        "    plt.gray()\n",
        "    ax.get_xaxis().set_visible(False)\n",
        "    ax.get_yaxis().set_visible(False)\n",
        "plt.show()"
      ],
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1440x144 with 9 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1440x144 with 9 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QtW_31_Q2IG6",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "reduced_x_train = reduced_x_train.reshape(reduced_x_train.shape[0], 14*14)\n",
        "reduced_x_test = reduced_x_test.reshape(reduced_x_test.shape[0], 14*14)\n",
        "x_train = x_train.reshape(x_train.shape[0], 28*28)\n",
        "x_test = x_test.reshape(x_test.shape[0], 28*28)"
      ],
      "execution_count": 37,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gUTop4Mg2z0T",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 84
        },
        "outputId": "fbb64c82-4115-433d-c28f-48eee3fb3fb1"
      },
      "source": [
        "print(reduced_x_train.shape)\n",
        "print(x_train.shape)\n",
        "print(reduced_x_test.shape)\n",
        "print(x_test.shape)"
      ],
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(60000, 196)\n",
            "(60000, 784)\n",
            "(10000, 196)\n",
            "(10000, 784)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Xqt71jtz2fjo",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "162e65a0-777b-49c0-a05b-25039e34ef89"
      },
      "source": [
        "# 모델 정의\n",
        "model = Sequential()\n",
        "model.add(Dense(128, activation='relu', input_dim=14*14))\n",
        "model.add(Dense(64, activation='relu'))\n",
        "model.add(Dense(32, activation='relu'))\n",
        "model.add(Dense(64, activation='relu'))\n",
        "model.add(Dense(128, activation='relu'))\n",
        "model.add(Dense(784, activation='sigmoid'))\n",
        "# model.add(Dense(784))\n",
        "\n",
        "model.compile(optimizer='adam', loss='mse')\n",
        "\n",
        "# 모델 학습\n",
        "model.fit(reduced_x_train, x_train, epochs=50, batch_size=256, shuffle=True, validation_split=0.1)"
      ],
      "execution_count": 46,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 54000 samples, validate on 6000 samples\n",
            "Epoch 1/50\n",
            "54000/54000 [==============================] - 2s 31us/sample - loss: 0.0606 - val_loss: 0.0345\n",
            "Epoch 2/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0287 - val_loss: 0.0247\n",
            "Epoch 3/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0230 - val_loss: 0.0209\n",
            "Epoch 4/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0200 - val_loss: 0.0187\n",
            "Epoch 5/50\n",
            "54000/54000 [==============================] - 1s 21us/sample - loss: 0.0183 - val_loss: 0.0174\n",
            "Epoch 6/50\n",
            "54000/54000 [==============================] - 1s 21us/sample - loss: 0.0171 - val_loss: 0.0164\n",
            "Epoch 7/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0160 - val_loss: 0.0153\n",
            "Epoch 8/50\n",
            "54000/54000 [==============================] - 1s 21us/sample - loss: 0.0152 - val_loss: 0.0148\n",
            "Epoch 9/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0146 - val_loss: 0.0144\n",
            "Epoch 10/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0141 - val_loss: 0.0136\n",
            "Epoch 11/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0135 - val_loss: 0.0132\n",
            "Epoch 12/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0132 - val_loss: 0.0129\n",
            "Epoch 13/50\n",
            "54000/54000 [==============================] - 1s 21us/sample - loss: 0.0128 - val_loss: 0.0126\n",
            "Epoch 14/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0125 - val_loss: 0.0123\n",
            "Epoch 15/50\n",
            "54000/54000 [==============================] - 1s 21us/sample - loss: 0.0122 - val_loss: 0.0121\n",
            "Epoch 16/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0120 - val_loss: 0.0120\n",
            "Epoch 17/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0117 - val_loss: 0.0115\n",
            "Epoch 18/50\n",
            "54000/54000 [==============================] - 1s 21us/sample - loss: 0.0114 - val_loss: 0.0113\n",
            "Epoch 19/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0112 - val_loss: 0.0111\n",
            "Epoch 20/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0110 - val_loss: 0.0109\n",
            "Epoch 21/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0109 - val_loss: 0.0109\n",
            "Epoch 22/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0107 - val_loss: 0.0107\n",
            "Epoch 23/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0106 - val_loss: 0.0106\n",
            "Epoch 24/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0105 - val_loss: 0.0105\n",
            "Epoch 25/50\n",
            "54000/54000 [==============================] - 1s 21us/sample - loss: 0.0104 - val_loss: 0.0103\n",
            "Epoch 26/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0103 - val_loss: 0.0103\n",
            "Epoch 27/50\n",
            "54000/54000 [==============================] - 1s 21us/sample - loss: 0.0102 - val_loss: 0.0101\n",
            "Epoch 28/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0100 - val_loss: 0.0101\n",
            "Epoch 29/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0100 - val_loss: 0.0100\n",
            "Epoch 30/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0099 - val_loss: 0.0099\n",
            "Epoch 31/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0098 - val_loss: 0.0098\n",
            "Epoch 32/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0097 - val_loss: 0.0097\n",
            "Epoch 33/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0096 - val_loss: 0.0098\n",
            "Epoch 34/50\n",
            "54000/54000 [==============================] - 1s 21us/sample - loss: 0.0096 - val_loss: 0.0097\n",
            "Epoch 35/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0095 - val_loss: 0.0096\n",
            "Epoch 36/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0094 - val_loss: 0.0095\n",
            "Epoch 37/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0094 - val_loss: 0.0095\n",
            "Epoch 38/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0093 - val_loss: 0.0094\n",
            "Epoch 39/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0093 - val_loss: 0.0096\n",
            "Epoch 40/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0092 - val_loss: 0.0093\n",
            "Epoch 41/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0092 - val_loss: 0.0093\n",
            "Epoch 42/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0091 - val_loss: 0.0092\n",
            "Epoch 43/50\n",
            "54000/54000 [==============================] - 1s 21us/sample - loss: 0.0091 - val_loss: 0.0092\n",
            "Epoch 44/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0090 - val_loss: 0.0091\n",
            "Epoch 45/50\n",
            "54000/54000 [==============================] - 1s 21us/sample - loss: 0.0090 - val_loss: 0.0091\n",
            "Epoch 46/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0089 - val_loss: 0.0090\n",
            "Epoch 47/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0089 - val_loss: 0.0090\n",
            "Epoch 48/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0088 - val_loss: 0.0090\n",
            "Epoch 49/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0088 - val_loss: 0.0090\n",
            "Epoch 50/50\n",
            "54000/54000 [==============================] - 1s 20us/sample - loss: 0.0088 - val_loss: 0.0089\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f069022bc50>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 46
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2y4hQ-VfFxF4",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 318
        },
        "outputId": "f0d5ccd2-0cdf-4edf-e43e-4d7096ee7030"
      },
      "source": [
        "# 결과 확인\n",
        "super_resolutioned = model.predict(reduced_x_test)\n",
        "\n",
        "n = 10\n",
        "plt.figure(figsize=(20, 6))\n",
        "for i in range(1, n):\n",
        "    # display original\n",
        "    ax = plt.subplot(3, n, i)\n",
        "    plt.imshow(x_test[i].reshape(28,28))\n",
        "    plt.gray()\n",
        "    ax.get_xaxis().set_visible(False)\n",
        "    ax.get_yaxis().set_visible(False)\n",
        "\n",
        "    # display reduced\n",
        "    ax = plt.subplot(3, n, i + n)\n",
        "    plt.imshow(reduced_x_test[i].reshape(14,14))\n",
        "    plt.gray()\n",
        "    ax.get_xaxis().set_visible(False)\n",
        "    ax.get_yaxis().set_visible(False)\n",
        "\n",
        "    # display super resolution\n",
        "    ax = plt.subplot(3, n, i + 2*n)\n",
        "    plt.imshow(super_resolutioned[i].reshape(28,28))\n",
        "    plt.gray()\n",
        "    ax.get_xaxis().set_visible(False)\n",
        "    ax.get_yaxis().set_visible(False)\n",
        "\n",
        "plt.show()"
      ],
      "execution_count": 47,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1440x432 with 27 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fTqVK4F6F1S2",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}