Facial_Key_Point_Detection.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Facial Key Point Detection.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "include_colab_link": true
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.6.6"
    },
    "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/pranaya-mathur/Deep-Learning-Projects/blob/master/Facial_Key_Point_Detection.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "I6gZkYly7e_u",
        "colab_type": "text"
      },
      "source": [
        "## 1. Data Exploration"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "87R0FgT57e_-",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "%matplotlib inline\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1-DSRrW07fAf",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "seed = 42\n",
        "rng = np.random.RandomState(seed)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "a-suh-Ym8HmA",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 72
        },
        "outputId": "395e273a-60f3-4dd5-ed93-ae0e59200cfe"
      },
      "source": [
        "!curl --header 'Host: storage.googleapis.com' --user-agent 'Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:67.0) Gecko/20100101 Firefox/67.0' --header 'Accept: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8' --header 'Accept-Language: en-US,en;q=0.5' --header 'Upgrade-Insecure-Requests: 1' 'https://storage.googleapis.com/kaggle-datasets/64052/125245/facial-key.zip?GoogleAccessId=web-data@kaggle-161607.iam.gserviceaccount.com&Expires=1563223567&Signature=U2Rah%2BCyc%2Bj4CJVWGYC9FmfU1uMnQ%2FfNWnPL1LsOmZifqWMOjFfVI9pfUDdPuqKawsWZi0pWCcX5WeL502xZNKHG9bUTyuVj8%2BWpJ90cjgqJegpLPwo3xNP3W34rLg2x87mNwI%2Flf6kacBvEWsYhvYJIpLoHiZXlfdSyltoinqlNx9dS8mSqJ99EG185Zt5Gk2p004Ip9hxUInT760PdvbQtGk3uHzaQz70w5KsiaqpiFHyqlTaOEW9Dn1p54p5t2u3ZgMVbDhAvmbAQem4I6xNz52S4%2B5he1xtf3Yt1oARj206BJTrz2fg3%2FCLAwNGfzQNvC3nriMrQG00iGb92xg%3D%3D' --output 'facial-key.zip'"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current\n",
            "                                 Dload  Upload   Total   Spent    Left  Speed\n",
            "100 29.2M  100 29.2M    0     0  32.3M      0 --:--:-- --:--:-- --:--:-- 32.2M\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "09v3qrsB8WZq",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import zipfile\n",
        "zip_ref = zipfile.ZipFile('facial-key.zip', 'r')\n",
        "zip_ref.extractall(\"data\")\n",
        "zip_ref.close()\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pYrXkqTu7fA8",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "data = pd.read_csv('data/facial_keypoint_identification/facial_keypoint_identification.csv')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "muLy1Jjc7fBO",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 206
        },
        "outputId": "9cb2ee02-05bc-41af-8351-c6e17820e8f1"
      },
      "source": [
        "data.head()"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>image_name</th>\n",
              "      <th>left_eye_center_x</th>\n",
              "      <th>left_eye_center_y</th>\n",
              "      <th>right_eye_center_x</th>\n",
              "      <th>right_eye_center_y</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>0.jpg</td>\n",
              "      <td>66</td>\n",
              "      <td>39</td>\n",
              "      <td>30</td>\n",
              "      <td>36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>1.jpg</td>\n",
              "      <td>64</td>\n",
              "      <td>34</td>\n",
              "      <td>29</td>\n",
              "      <td>33</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>2.jpg</td>\n",
              "      <td>65</td>\n",
              "      <td>34</td>\n",
              "      <td>30</td>\n",
              "      <td>34</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>3.jpg</td>\n",
              "      <td>65</td>\n",
              "      <td>37</td>\n",
              "      <td>32</td>\n",
              "      <td>37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>4.jpg</td>\n",
              "      <td>66</td>\n",
              "      <td>39</td>\n",
              "      <td>32</td>\n",
              "      <td>38</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "  image_name  left_eye_center_x  ...  right_eye_center_x  right_eye_center_y\n",
              "0      0.jpg                 66  ...                  30                  36\n",
              "1      1.jpg                 64  ...                  29                  33\n",
              "2      2.jpg                 65  ...                  30                  34\n",
              "3      3.jpg                 65  ...                  32                  37\n",
              "4      4.jpg                 66  ...                  32                  38\n",
              "\n",
              "[5 rows x 5 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WLHSkC9f7fBq",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 287
        },
        "outputId": "cbb10e6a-56c3-4311-f5fb-5c2cb5ebd49e"
      },
      "source": [
        "# get random index of data\n",
        "idx = rng.choice(data.index)\n",
        "\n",
        "# read image\n",
        "img = plt.imread('data/facial_keypoint_identification/images/' + data.image_name.iloc[idx])\n",
        "\n",
        "# create an empty figure to plot the image and target\n",
        "fig = plt.figure()\n",
        "\n",
        "# plot image\n",
        "plt.imshow(img)\n",
        "\n",
        "# plot target\n",
        "plt.scatter(data.left_eye_center_x.iloc[idx], data.left_eye_center_y.iloc[idx], marker='x', c='r')\n",
        "plt.scatter(data.right_eye_center_x.iloc[idx], data.right_eye_center_y.iloc[idx], marker='x', c='r')\n",
        "\n",
        "# and print the shape of image\n",
        "print('Shape of original image is', img.shape)\n",
        "\n",
        "# finally show the figure\n",
        "plt.show()"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Shape of original image is (96, 96, 3)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD8CAYAAABXXhlaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nOUT/4Xd0zCH6D7+jYw7x/wF1Clyz292ZdwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "p88Q_DoX7fCD",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn.metrics import mean_squared_error"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YeVRFL8E7fCS",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "923dc3c7-ea58-437f-e432-7e3fedf5814f"
      },
      "source": [
        "error_left_x = mean_squared_error(data.left_eye_center_x.values, np.full(data.left_eye_center_x.values.shape, data.left_eye_center_x.mean()))\n",
        "error_left_y = mean_squared_error(data.left_eye_center_y.values, np.full(data.left_eye_center_y.values.shape, data.left_eye_center_y.mean()))\n",
        "error_right_x = mean_squared_error(data.right_eye_center_x.values, np.full(data.right_eye_center_x.values.shape, data.right_eye_center_x.mean()))\n",
        "error_right_y = mean_squared_error(data.right_eye_center_y.values, np.full(data.right_eye_center_y.values.shape, data.right_eye_center_y.mean()))\n",
        "\n",
        "error = (error_left_x*error_left_x + error_left_y*error_left_y + error_right_x*error_right_x + error_right_y*error_right_y ) / 4\n",
        "error"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "104.5789947402131"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9xQA0mac7fCm",
        "colab_type": "text"
      },
      "source": [
        "## 2. Data Loading and Preprocessing"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oFEgzHOw7fCs",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "c2f3b079-1a68-4a8e-bac2-ea1098054478"
      },
      "source": [
        "from keras.preprocessing import image\n",
        "from keras.applications.vgg16 import preprocess_input\n",
        "from sklearn.model_selection import train_test_split"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9KSLE3aW7fDK",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "images = []\n",
        "for img in data.image_name:\n",
        "    img = image.load_img('data/facial_keypoint_identification/images/' + img, target_size=(224, 224, 3))\n",
        "    img = image.img_to_array(img)\n",
        "    images.append(img)\n",
        "    \n",
        "images=np.array(images)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "66OcRZe97fDa",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "26a286b1-7620-401b-86e8-f83f822c0445"
      },
      "source": [
        "images.shape"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(7033, 224, 224, 3)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ff_XkHOg7fDt",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "X=preprocess_input(images, mode='tf')\n",
        "y = data.iloc[:, 1:].values"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "V4dPJbcc7fEA",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "X_train, X_valid, Y_train, Y_valid=train_test_split(X,y,test_size=0.3, random_state=42)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kl1cV4su7fEQ",
        "colab_type": "text"
      },
      "source": [
        "## 3. Model Building"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PSJqPchg7fEU",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from keras.applications.vgg16 import VGG16\n",
        "\n",
        "from keras.models import Sequential\n",
        "from keras.optimizers import Adam\n",
        "from keras.layers import Dense, Dropout, InputLayer\n",
        "from keras.callbacks import EarlyStopping"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NUg3TeKe7fEu",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 313
        },
        "outputId": "eb7a54b1-dc37-4cfe-8672-21d1d20496c3"
      },
      "source": [
        "base_model = VGG16(include_top=False, weights='imagenet', input_shape=(224, 224, 3))"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING: Logging before flag parsing goes to stderr.\n",
            "W0712 20:51:09.294125 140272655267712 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:74: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.\n",
            "\n",
            "W0712 20:51:09.331834 140272655267712 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:517: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n",
            "\n",
            "W0712 20:51:09.338847 140272655267712 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4138: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.\n",
            "\n",
            "W0712 20:51:09.378549 140272655267712 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3976: The name tf.nn.max_pool is deprecated. Please use tf.nn.max_pool2d instead.\n",
            "\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://github.com/fchollet/deep-learning-models/releases/download/v0.1/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5\n",
            "58892288/58889256 [==============================] - 1s 0us/step\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "W0712 20:51:10.399283 140272655267712 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:174: The name tf.get_default_session is deprecated. Please use tf.compat.v1.get_default_session instead.\n",
            "\n",
            "W0712 20:51:10.400751 140272655267712 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:181: The name tf.ConfigProto is deprecated. Please use tf.compat.v1.ConfigProto instead.\n",
            "\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Bm3fmZZC7fE5",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Extracting features\n",
        "\n",
        "features_train = base_model.predict(X_train)\n",
        "features_valid = base_model.predict(X_valid)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NJBuil7P7fFx",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "6c36f2ea-905b-4316-b6c4-ca2d59215572"
      },
      "source": [
        "features_train.max()"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "10.810992"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pLx-XCP57fG-",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "max_val = features_train.max()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UaDXYrke7fHi",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "features_train /= max_val\n",
        "features_valid /= max_val"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aV5SRcnv7fH3",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "features_train = features_train.reshape(features_train.shape[0], 7*7*512)\n",
        "features_valid = features_valid.reshape(features_valid.shape[0], 7*7*512)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": true,
        "id": "FUxLaT3a7fIq",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 129
        },
        "outputId": "e7075dd4-0829-4bb1-da89-48e1f650f6f1"
      },
      "source": [
        "model=Sequential()\n",
        "model.add(InputLayer((7*7*512, )))\n",
        "model.add(Dense(1024, activation='relu'))\n",
        "model.add(Dropout(0.3))\n",
        "model.add(Dense(units=4, activation='linear'))\n",
        "\n",
        "adam = Adam(lr=1e-3)\n",
        "model.compile(loss='mean_squared_error', optimizer=adam)"
      ],
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "W0712 20:52:23.945881 140272655267712 deprecation.py:506] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
            "W0712 20:52:23.988918 140272655267712 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/optimizers.py:790: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n",
            "\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eoxWCvHQ7fJ6",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "estop = EarlyStopping(patience=10, mode='min', min_delta=0.001, monitor='val_loss')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": true,
        "id": "jQWBu_gG7fKo",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "6d8e8cc6-cf43-494c-e070-e01d9251a7ed"
      },
      "source": [
        "model.fit(features_train, Y_train, epochs=200, batch_size=512, callbacks=[estop] ,validation_data=(features_valid,Y_valid))"
      ],
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 4923 samples, validate on 2110 samples\n",
            "Epoch 1/200\n",
            "4923/4923 [==============================] - 3s 567us/step - loss: 1121.2657 - val_loss: 148.9870\n",
            "Epoch 2/200\n",
            "4923/4923 [==============================] - 1s 154us/step - loss: 142.3511 - val_loss: 167.4066\n",
            "Epoch 3/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 91.2469 - val_loss: 54.3718\n",
            "Epoch 4/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 55.3768 - val_loss: 34.5272\n",
            "Epoch 5/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 34.7124 - val_loss: 24.4463\n",
            "Epoch 6/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 28.5424 - val_loss: 18.9676\n",
            "Epoch 7/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 23.5424 - val_loss: 16.6473\n",
            "Epoch 8/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 21.3292 - val_loss: 14.9615\n",
            "Epoch 9/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 19.6904 - val_loss: 13.7437\n",
            "Epoch 10/200\n",
            "4923/4923 [==============================] - 1s 141us/step - loss: 18.8374 - val_loss: 12.9043\n",
            "Epoch 11/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 17.9318 - val_loss: 12.2516\n",
            "Epoch 12/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 17.2019 - val_loss: 11.6710\n",
            "Epoch 13/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 16.5702 - val_loss: 11.1813\n",
            "Epoch 14/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 16.2170 - val_loss: 10.7532\n",
            "Epoch 15/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 15.4303 - val_loss: 10.3577\n",
            "Epoch 16/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 14.9570 - val_loss: 10.0351\n",
            "Epoch 17/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 14.7613 - val_loss: 9.7264\n",
            "Epoch 18/200\n",
            "4923/4923 [==============================] - 1s 142us/step - loss: 14.2163 - val_loss: 9.4175\n",
            "Epoch 19/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 13.8814 - val_loss: 9.2060\n",
            "Epoch 20/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 13.5717 - val_loss: 8.9414\n",
            "Epoch 21/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 13.2053 - val_loss: 8.7829\n",
            "Epoch 22/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 12.8915 - val_loss: 8.5411\n",
            "Epoch 23/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 12.5077 - val_loss: 8.3663\n",
            "Epoch 24/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 12.1066 - val_loss: 8.1840\n",
            "Epoch 25/200\n",
            "4923/4923 [==============================] - 1s 142us/step - loss: 12.0282 - val_loss: 8.0308\n",
            "Epoch 26/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 11.7461 - val_loss: 7.8891\n",
            "Epoch 27/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 11.5244 - val_loss: 7.7407\n",
            "Epoch 28/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 11.2685 - val_loss: 7.6456\n",
            "Epoch 29/200\n",
            "4923/4923 [==============================] - 1s 147us/step - loss: 10.9319 - val_loss: 7.4912\n",
            "Epoch 30/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 10.9115 - val_loss: 7.3889\n",
            "Epoch 31/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 10.7089 - val_loss: 7.2855\n",
            "Epoch 32/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 10.3874 - val_loss: 7.1804\n",
            "Epoch 33/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 10.2160 - val_loss: 7.0921\n",
            "Epoch 34/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 10.2732 - val_loss: 7.0284\n",
            "Epoch 35/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 9.9556 - val_loss: 6.9169\n",
            "Epoch 36/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 9.8376 - val_loss: 6.8440\n",
            "Epoch 37/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 9.5077 - val_loss: 6.7523\n",
            "Epoch 38/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 9.3890 - val_loss: 6.6978\n",
            "Epoch 39/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 9.3339 - val_loss: 6.6111\n",
            "Epoch 40/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 9.2140 - val_loss: 6.5550\n",
            "Epoch 41/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 8.9843 - val_loss: 6.4956\n",
            "Epoch 42/200\n",
            "4923/4923 [==============================] - 1s 178us/step - loss: 9.1004 - val_loss: 6.4412\n",
            "Epoch 43/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 8.8374 - val_loss: 6.3787\n",
            "Epoch 44/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 8.6369 - val_loss: 6.3091\n",
            "Epoch 45/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 8.7059 - val_loss: 6.2541\n",
            "Epoch 46/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 8.4869 - val_loss: 6.2254\n",
            "Epoch 47/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 8.4095 - val_loss: 6.1644\n",
            "Epoch 48/200\n",
            "4923/4923 [==============================] - 1s 142us/step - loss: 8.3269 - val_loss: 6.1442\n",
            "Epoch 49/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 8.2039 - val_loss: 6.0749\n",
            "Epoch 50/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 8.0239 - val_loss: 6.1005\n",
            "Epoch 51/200\n",
            "4923/4923 [==============================] - 1s 142us/step - loss: 7.9561 - val_loss: 6.0211\n",
            "Epoch 52/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 8.0369 - val_loss: 5.9553\n",
            "Epoch 53/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 7.7704 - val_loss: 5.9259\n",
            "Epoch 54/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 7.8184 - val_loss: 5.8972\n",
            "Epoch 55/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 7.5866 - val_loss: 5.8371\n",
            "Epoch 56/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 7.5494 - val_loss: 5.8111\n",
            "Epoch 57/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 7.5520 - val_loss: 5.7607\n",
            "Epoch 58/200\n",
            "4923/4923 [==============================] - 1s 147us/step - loss: 7.3424 - val_loss: 5.7413\n",
            "Epoch 59/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 7.2521 - val_loss: 5.7060\n",
            "Epoch 60/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 7.2078 - val_loss: 5.6942\n",
            "Epoch 61/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 7.1796 - val_loss: 5.6380\n",
            "Epoch 62/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 7.3007 - val_loss: 5.6373\n",
            "Epoch 63/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 7.1329 - val_loss: 5.6499\n",
            "Epoch 64/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 7.0790 - val_loss: 5.5671\n",
            "Epoch 65/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 6.9762 - val_loss: 5.5494\n",
            "Epoch 66/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 6.9767 - val_loss: 5.5201\n",
            "Epoch 67/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 6.8096 - val_loss: 5.5264\n",
            "Epoch 68/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 6.7408 - val_loss: 5.4933\n",
            "Epoch 69/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 6.7342 - val_loss: 5.4478\n",
            "Epoch 70/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 6.6506 - val_loss: 5.4337\n",
            "Epoch 71/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 6.6062 - val_loss: 5.4083\n",
            "Epoch 72/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 6.3859 - val_loss: 5.3885\n",
            "Epoch 73/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 6.5194 - val_loss: 5.3677\n",
            "Epoch 74/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 6.4679 - val_loss: 5.3487\n",
            "Epoch 75/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 6.3410 - val_loss: 5.3470\n",
            "Epoch 76/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 6.4041 - val_loss: 5.3269\n",
            "Epoch 77/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 6.2537 - val_loss: 5.3470\n",
            "Epoch 78/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 6.2803 - val_loss: 5.3522\n",
            "Epoch 79/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 6.1407 - val_loss: 5.3004\n",
            "Epoch 80/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 6.2097 - val_loss: 5.2554\n",
            "Epoch 81/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 5.9093 - val_loss: 5.2298\n",
            "Epoch 82/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 6.0022 - val_loss: 5.2383\n",
            "Epoch 83/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 6.0764 - val_loss: 5.2261\n",
            "Epoch 84/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 5.9557 - val_loss: 5.2545\n",
            "Epoch 85/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 5.9571 - val_loss: 5.2054\n",
            "Epoch 86/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 5.8557 - val_loss: 5.1481\n",
            "Epoch 87/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 5.7996 - val_loss: 5.1490\n",
            "Epoch 88/200\n",
            "4923/4923 [==============================] - 1s 148us/step - loss: 5.6851 - val_loss: 5.1347\n",
            "Epoch 89/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 5.7533 - val_loss: 5.1361\n",
            "Epoch 90/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 5.6915 - val_loss: 5.1342\n",
            "Epoch 91/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 5.6431 - val_loss: 5.0991\n",
            "Epoch 92/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 5.6181 - val_loss: 5.0903\n",
            "Epoch 93/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 5.6916 - val_loss: 5.0822\n",
            "Epoch 94/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 5.5721 - val_loss: 5.1544\n",
            "Epoch 95/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 5.5285 - val_loss: 5.1305\n",
            "Epoch 96/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 5.5671 - val_loss: 5.1029\n",
            "Epoch 97/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 5.5213 - val_loss: 5.0775\n",
            "Epoch 98/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 5.4245 - val_loss: 5.0449\n",
            "Epoch 99/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 5.3515 - val_loss: 5.0459\n",
            "Epoch 100/200\n",
            "4923/4923 [==============================] - 1s 142us/step - loss: 5.4209 - val_loss: 5.0240\n",
            "Epoch 101/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 5.2022 - val_loss: 5.0135\n",
            "Epoch 102/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 5.3406 - val_loss: 4.9938\n",
            "Epoch 103/200\n",
            "4923/4923 [==============================] - 1s 147us/step - loss: 5.3161 - val_loss: 4.9941\n",
            "Epoch 104/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 5.1697 - val_loss: 5.0357\n",
            "Epoch 105/200\n",
            "4923/4923 [==============================] - 1s 147us/step - loss: 5.1467 - val_loss: 5.0503\n",
            "Epoch 106/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 5.2133 - val_loss: 4.9645\n",
            "Epoch 107/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 5.1852 - val_loss: 4.9714\n",
            "Epoch 108/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 5.1735 - val_loss: 4.9621\n",
            "Epoch 109/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 5.1073 - val_loss: 4.9997\n",
            "Epoch 110/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 5.0933 - val_loss: 4.9447\n",
            "Epoch 111/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 5.0992 - val_loss: 4.9763\n",
            "Epoch 112/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 5.1278 - val_loss: 4.9356\n",
            "Epoch 113/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 5.1045 - val_loss: 5.0233\n",
            "Epoch 114/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 5.1270 - val_loss: 4.9237\n",
            "Epoch 115/200\n",
            "4923/4923 [==============================] - 1s 147us/step - loss: 5.0894 - val_loss: 4.9247\n",
            "Epoch 116/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.8699 - val_loss: 4.9164\n",
            "Epoch 117/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.9058 - val_loss: 4.9535\n",
            "Epoch 118/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 5.0211 - val_loss: 4.9435\n",
            "Epoch 119/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 4.8386 - val_loss: 4.8984\n",
            "Epoch 120/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.7595 - val_loss: 4.8898\n",
            "Epoch 121/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.8434 - val_loss: 4.8852\n",
            "Epoch 122/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.7940 - val_loss: 4.9029\n",
            "Epoch 123/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.7577 - val_loss: 4.9420\n",
            "Epoch 124/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 4.7879 - val_loss: 4.8887\n",
            "Epoch 125/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.7734 - val_loss: 4.9521\n",
            "Epoch 126/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 4.7318 - val_loss: 4.9135\n",
            "Epoch 127/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.6885 - val_loss: 4.8717\n",
            "Epoch 128/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 4.6270 - val_loss: 4.8818\n",
            "Epoch 129/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.6074 - val_loss: 4.9469\n",
            "Epoch 130/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 4.6679 - val_loss: 4.9079\n",
            "Epoch 131/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.7633 - val_loss: 4.8859\n",
            "Epoch 132/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.5410 - val_loss: 4.8986\n",
            "Epoch 133/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.6300 - val_loss: 4.8642\n",
            "Epoch 134/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.6093 - val_loss: 4.8433\n",
            "Epoch 135/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.3546 - val_loss: 4.8337\n",
            "Epoch 136/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.4911 - val_loss: 4.8438\n",
            "Epoch 137/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.5281 - val_loss: 4.8750\n",
            "Epoch 138/200\n",
            "4923/4923 [==============================] - 1s 147us/step - loss: 4.4792 - val_loss: 4.8381\n",
            "Epoch 139/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.4205 - val_loss: 4.9179\n",
            "Epoch 140/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.3862 - val_loss: 4.8499\n",
            "Epoch 141/200\n",
            "4923/4923 [==============================] - 1s 147us/step - loss: 4.4602 - val_loss: 4.8493\n",
            "Epoch 142/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 4.3664 - val_loss: 4.8389\n",
            "Epoch 143/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.4340 - val_loss: 4.8164\n",
            "Epoch 144/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.3794 - val_loss: 4.9428\n",
            "Epoch 145/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 4.3807 - val_loss: 5.0178\n",
            "Epoch 146/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.3930 - val_loss: 4.8822\n",
            "Epoch 147/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.2803 - val_loss: 4.8514\n",
            "Epoch 148/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 4.3525 - val_loss: 4.8540\n",
            "Epoch 149/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.3721 - val_loss: 4.8535\n",
            "Epoch 150/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 4.2204 - val_loss: 4.8158\n",
            "Epoch 151/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 4.3017 - val_loss: 4.8101\n",
            "Epoch 152/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 4.2529 - val_loss: 4.8011\n",
            "Epoch 153/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 4.2560 - val_loss: 4.8660\n",
            "Epoch 154/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 4.1362 - val_loss: 4.8062\n",
            "Epoch 155/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 4.1940 - val_loss: 4.9045\n",
            "Epoch 156/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 4.1353 - val_loss: 4.8050\n",
            "Epoch 157/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.2437 - val_loss: 4.8320\n",
            "Epoch 158/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.1664 - val_loss: 4.7888\n",
            "Epoch 159/200\n",
            "4923/4923 [==============================] - 1s 142us/step - loss: 4.1441 - val_loss: 4.8466\n",
            "Epoch 160/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.1392 - val_loss: 4.7830\n",
            "Epoch 161/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 4.1231 - val_loss: 4.9045\n",
            "Epoch 162/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.1498 - val_loss: 4.8065\n",
            "Epoch 163/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 4.0930 - val_loss: 4.9739\n",
            "Epoch 164/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.1406 - val_loss: 4.7967\n",
            "Epoch 165/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.0837 - val_loss: 4.7637\n",
            "Epoch 166/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.1295 - val_loss: 4.7860\n",
            "Epoch 167/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.0542 - val_loss: 4.8846\n",
            "Epoch 168/200\n",
            "4923/4923 [==============================] - 1s 146us/step - loss: 4.0580 - val_loss: 4.7723\n",
            "Epoch 169/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.0091 - val_loss: 4.7898\n",
            "Epoch 170/200\n",
            "4923/4923 [==============================] - 1s 145us/step - loss: 4.0643 - val_loss: 4.8166\n",
            "Epoch 171/200\n",
            "4923/4923 [==============================] - 1s 142us/step - loss: 4.0671 - val_loss: 4.7821\n",
            "Epoch 172/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 4.0917 - val_loss: 4.7920\n",
            "Epoch 173/200\n",
            "4923/4923 [==============================] - 1s 144us/step - loss: 3.8841 - val_loss: 4.7786\n",
            "Epoch 174/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 3.9306 - val_loss: 4.7835\n",
            "Epoch 175/200\n",
            "4923/4923 [==============================] - 1s 143us/step - loss: 3.9591 - val_loss: 4.7685\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7f93603f9128>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 28
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "06IuQcjO7fLq",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 269
        },
        "outputId": "e336d87e-0f79-4059-8edf-1a4b14bc5fee"
      },
      "source": [
        "# get predictions\n",
        "predictions = model.predict(features_valid)\n",
        "\n",
        "# extract validation images\n",
        "_, valid_images, _, _ = train_test_split(data.image_name,y,test_size=0.3, random_state=42)\n",
        "\n",
        "# get random validation image index\n",
        "idx = rng.choice(range(len(valid_images)))\n",
        "\n",
        "# get the corresponding image name\n",
        "image_name = valid_images.iloc[idx]\n",
        "\n",
        "# get the actual targets\n",
        "left_eye_center_x = data.loc[data.image_name == image_name, 'left_eye_center_x']\n",
        "left_eye_center_y = data.loc[data.image_name == image_name, 'left_eye_center_y']\n",
        "right_eye_center_x = data.loc[data.image_name == image_name, 'right_eye_center_x']\n",
        "right_eye_center_y = data.loc[data.image_name == image_name, 'right_eye_center_y']\n",
        "\n",
        "# get the corresponding predictions\n",
        "predicted_left_eye_center_x, predicted_left_eye_center_y, predicted_right_eye_center_x, predicted_right_eye_center_y = predictions[idx]\n",
        "\n",
        "# read the image\n",
        "img = plt.imread('data/facial_keypoint_identification/images/' + image_name)\n",
        "\n",
        "# plot empty figure\n",
        "fig = plt.figure()\n",
        "\n",
        "# plot image\n",
        "plt.imshow(img,cmap='gray')\n",
        "\n",
        "# plot actual targets\n",
        "plt.scatter(left_eye_center_x, left_eye_center_y, marker='x', s=50, c='r')\n",
        "plt.scatter(right_eye_center_x, right_eye_center_y, marker='x', s=50, c='r')\n",
        "\n",
        "# plot predictions\n",
        "plt.scatter(predicted_left_eye_center_x, predicted_left_eye_center_y, marker='o', s=50, c='w')\n",
        "plt.scatter(predicted_right_eye_center_x, predicted_right_eye_center_y, marker='o', s=50, c='w')\n",
        "\n",
        "# show the figure\n",
        "plt.show()"
      ],
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD8CAYAAABXXhlaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nfkPDFKL98BsaphD/H3ILVmBrUj2eAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8zlc0BPt7fL5",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}