Age_regression_solution.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "oa10OpnEvNm9"
   },
   "source": [
    "# Deep learning programming I-B: Age Regression\n",
    "Felix Wiewel, Institute of Signal Processing and System Theory, University of Stuttgart, 24.04.2020\n",
    "\n",
    "## Introduction\n",
    "This exercise is the second part of the regression homework in the course Deep Learning. While we covered a very basic regression task on synthetic data in the last part, we will now turn to the task of age regression. The task is to infer the age of a person given a potrait picture of him or her. Intuatively this task seems to be easy at first, you are shown a potrait of a person and need to estimate his or her age. But implementing such an estimator on a computer is rather challenging. If you want to use a model based machine learning algorithm, you would need some sort of model or set of rules, which maps the raw pixel values of an image to an estimated age. This would require a lot of expert knowledge and feature design. Fortunately we can use neural networks to learn such a mapping directly from annotated examples, i.e. potraits labeled with the age of the person in the picture. In this way we just need enough examples and don't have to design the probably very complex mapping from pixels to age ourselves. We will build on the mathematical fomulation from the last exercise and take a sligthly different approach for the implementation. So it is recommended to work through the first part of the exercise before starting with this part.\n",
    "\n",
    "## GPU support\n",
    "In order to speed up calculations with Tensorflow, we need to change the runtime type of this notebook to GPU. For this click on \"Runtime\" in the top left menu and select \"Change runtime type\". Then choose \"GPU\" in the drop down list under \"Hardware accelerator\". This will enable Tensorflow to execute calculations on a GPU provided by Google Colab.\n",
    "\n",
    "## Data Set\n",
    "The data set we use in this exercise is called UTKFaces (https://susanqq.github.io/UTKFace/). It contains over $20000$ images of people with ages from the interval $\\left[0,116\\right]$. Although the individual images are labeled with the age, gender and ethnicity, we will only use the age annotation in this exercise. The images in this data set are available in two different sets. The \"In-the-wild Faces\" set contains images in their original size and orientation and the \"Aligned&Cropped Faces\" contains all images with the face aligned in the middle and the size cropped to $200\\times 200$ pixels. Since the aligned and cropped set is simpler to work with and requires less memory, we will use it in this exercise. But before we can use the data set, we have to download it. This is done with the following commands."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 605
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 68501,
     "status": "ok",
     "timestamp": 1562140793562,
     "user": {
      "displayName": "Felix Wiewel",
      "photoUrl": "https://lh3.googleusercontent.com/-XBgpqQQzeMU/AAAAAAAAAAI/AAAAAAAAAZY/12qzl-EFXfM/s64/photo.jpg",
      "userId": "11509421729312238102"
     },
     "user_tz": -120
    },
    "id": "ZaWLfEhUu4QM",
    "outputId": "6f2727b6-8443-47f3-af90-2848d6b90341"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mkdir: cannot create directory ‘data’: File exists\n",
      "/bin/sh: 1: gdown: not found\n",
      "mv: cannot stat 'UTKFace.tar.gz': No such file or directory\n"
     ]
    }
   ],
   "source": [
    "!mkdir data\n",
    "!gdown https://drive.google.com/uc?id=0BxYys69jI14kYVM3aVhKS1VhRUk\n",
    "!mv UTKFace.tar.gz data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "o1pDCzZ6Dk2t"
   },
   "source": [
    "With these commands we created a folder named \"data\" and downloaded the data set into it. The data set is contained in an archive, which needs to be extracted before we can work with it. Since we will need to load individual images in the coming exercise, it makes sense to first extract the data and then retreive a list with all extracted files. For this we first import all necessary packages and then define a function, which extracts the data and returns a list with all available file names and since the labels are encoded in the file names we directly extract a corresponding list of labels."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "9Rysv6DCDkft"
   },
   "outputs": [],
   "source": [
    "import tarfile\n",
    "import os\n",
    "import glob\n",
    "import tensorflow as tf\n",
    "import tensorflow.keras as k\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "\n",
    "def load_file_names():\n",
    "    # Check if data has been extracted and if not extract it\n",
    "    if (os.path.isdir(\"./data/UTKFace\")):\n",
    "        print(\"Data set already extracted...\")\n",
    "    else:\n",
    "        print(\"Extracting data set...\")\n",
    "        tar = tarfile.open(\"./data/UTKFace.tar.gz\")\n",
    "        tar.extractall(\"./data\")\n",
    "        tar.close()\n",
    "\n",
    "    # Get a list of all files in data set\n",
    "    files = glob.glob(\"./data/UTKFace/*.jpg\")\n",
    "    labels = [int(f_name.split(\"/\")[-1].split(\"_\")[0]) for f_name in files]\n",
    "    return files, labels"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "XUCJEBM1EY-z"
   },
   "source": [
    "We can now use this function to plot an individual image and the age of the person in that image.."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 299
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 79155,
     "status": "ok",
     "timestamp": 1562140804237,
     "user": {
      "displayName": "Felix Wiewel",
      "photoUrl": "https://lh3.googleusercontent.com/-XBgpqQQzeMU/AAAAAAAAAAI/AAAAAAAAAZY/12qzl-EFXfM/s64/photo.jpg",
      "userId": "11509421729312238102"
     },
     "user_tz": -120
    },
    "id": "lhV-eakrE0XB",
    "outputId": "4acda405-ba6a-40ac-e43d-aee461378174"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Data set already extracted...\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "files, labels = load_file_names()\n",
    "img = plt.imread(files[0])\n",
    "plt.imshow(img)\n",
    "plt.title(\"Age: {}\".format(labels[0]))\n",
    "plt.axis(\"off\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "dYW0N607dHFk"
   },
   "source": [
    "We should also check the distribution of the labels. For this we will print the unique labels contained in our data set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 127
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 79146,
     "status": "ok",
     "timestamp": 1562140804242,
     "user": {
      "displayName": "Felix Wiewel",
      "photoUrl": "https://lh3.googleusercontent.com/-XBgpqQQzeMU/AAAAAAAAAAI/AAAAAAAAAZY/12qzl-EFXfM/s64/photo.jpg",
      "userId": "11509421729312238102"
     },
     "user_tz": -120
    },
    "id": "fLOkahM9dTxI",
    "outputId": "91778b77-8488-4672-9911-415d7ce6c8e2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[  1   2   3   4   5   6   7   8   9  10  11  12  13  14  15  16  17  18\n",
      "  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36\n",
      "  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54\n",
      "  55  56  57  58  59  60  61  62  63  64  65  66  67  68  69  70  71  72\n",
      "  73  74  75  76  77  78  79  80  81  82  83  84  85  86  87  88  89  90\n",
      "  91  92  93  95  96  99 100 101 103 105 110 111 115 116]\n"
     ]
    }
   ],
   "source": [
    "print(np.unique(labels))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "2VkgukOnaDI9"
   },
   "source": [
    "As we can see, the data set contains only integer values as the age label. With such a distribution of labels you could also try to solve the problem of mapping an input image to an age by classifiaction. This is a valid approach but not the one we are interested in. We want to take an image of a person and regress his or her age exactly. Since the potraits in our training data set were certainly not taken in the exact moment that person's age was exactly an integer, we can conclude that the labeling is not well suited for our problem. But since these are the only labels we get for this data set, we have to use them and accept the slightly incorrect labels. We can also think of this discrepancy between the exact age of the person when the photo was taken and the assigned label as noise, i.e. we are given noisy labels.\n",
    "\n",
    "In this exercise we will use a slightly different way to define our model when compared with the previous part. This time we will derive a new model class from the Model class provided by Keras. This comes with several advantages. First, we can use predefined layer classes from Keras in order to build our model. See the [documentation](https://www.tensorflow.org/versions/r2.2/api_docs/python/tf/keras/layers) for a list of all available layers and detailed information on how to use them. Second, we don't need to manage the variables which store the weights of our model manually. Keras takes care of this and provides a list of all trainable variables of the model. Third, we can use the fit and fit_generator function of a Keras model in order to easily train our model. These functions offer a simple interface to implement a lot of different ways for training. But since sometimes one needs to implement a custom training loop, we will not use this function yet but rather implement our own training loop. Now your task is to define a suitable model architecture for the regression of a scalar value like the age of a person from an image."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "v4Kd0bwDcgKH"
   },
   "outputs": [],
   "source": [
    "class MyModel(k.Model):\n",
    "    def __init__(self):\n",
    "        super(MyModel, self).__init__()\n",
    "        self.conv0 = k.layers.Conv2D(8, 5, 4, activation=\"relu\")\n",
    "        self.conv1 = k.layers.Conv2D(16, 5, 4, activation=\"relu\")\n",
    "        self.flatten = k.layers.Flatten()\n",
    "        self.dropout = k.layers.Dropout(0.2)\n",
    "        self.dense0 = k.layers.Dense(128, activation=\"relu\")\n",
    "        self.dense1 = k.layers.Dense(64, activation=\"relu\")\n",
    "        self.dense2 = k.layers.Dense(1, activation=\"relu\")\n",
    "\n",
    "    def call(self, inputs, training=False):\n",
    "        output = self.conv0(inputs)\n",
    "        output = self.conv1(output)\n",
    "        output = self.flatten(output)\n",
    "        output = self.dropout(output, training)\n",
    "        output = self.dense0(output)\n",
    "        output = self.dense1(output)\n",
    "        output = self.dense2(output)\n",
    "        return output"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "qjT0wAvsfrWC"
   },
   "source": [
    "Notice that in the previous part we defined a \\__call__ function and now define a call function. This call function is required by Keras but we can still predict with the model just by calling it on an input tensor. For training the model we again need to define some hyperparameters."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "VWErNUSCfxYu"
   },
   "outputs": [],
   "source": [
    "N_epochs = 20\n",
    "learning_rate = 0.001\n",
    "batch_size = 64\n",
    "N_training_examples = 20000\n",
    "N_validation_examples = 4*batch_size\n",
    "\n",
    "N_parallel_iterations = 4\n",
    "N_prefetch = 8\n",
    "N_shuffle_buffer = 20000"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "GWP_R8I9f5oy"
   },
   "source": [
    "We will train the model for $20$ epochs with a batch size of $64$ and a learning rate of $0.001$.  We will use $20000$ examples for training $4$ batches for validation and the remaining data for testing of the model. We also define some parameters, which are used to create the Dataset objects. With these definitions we can build the data set objects used for training, validation and testing."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "tDmQ11FQgeGP"
   },
   "outputs": [],
   "source": [
    "def parse_func(filename, label):\n",
    "    image_string = tf.io.read_file(filename)\n",
    "    image_decoded = tf.io.decode_jpeg(image_string)\n",
    "    image = tf.cast(image_decoded, tf.float32)/255.0\n",
    "    label = tf.expand_dims(tf.cast(label, tf.float32), axis=-1)\n",
    "    return image, label\n",
    "\n",
    "def build_dataset(files, labels, batch_size):\n",
    "    # Create tf data set\n",
    "    ds = tf.data.Dataset.from_tensor_slices((files, labels))\n",
    "    ds = ds.shuffle(N_shuffle_buffer)\n",
    "    ds = ds.map(parse_func, N_parallel_iterations)\n",
    "    ds = ds.batch(batch_size).repeat().prefetch(N_prefetch)\n",
    "    return ds\n",
    "\n",
    "shuffle_idx = np.arange(len(files))\n",
    "np.random.shuffle(shuffle_idx)\n",
    "files = [files[i] for i in shuffle_idx]\n",
    "labels = [labels[i] for i in shuffle_idx]\n",
    "train_ds = build_dataset(files[0:N_training_examples],\n",
    "                                     labels[0:N_training_examples], batch_size)\n",
    "validation_ds = build_dataset(files[N_training_examples:N_training_examples+N_validation_examples],\n",
    "                                          labels[N_training_examples:N_training_examples+N_validation_examples],\n",
    "                                          batch_size)\n",
    "test_ds = build_dataset(files[N_training_examples+N_validation_examples:],\n",
    "                                    labels[N_training_examples+N_validation_examples:], batch_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "s_oV4_VThr5f"
   },
   "source": [
    "Since the we loaded the file names of individual images and their corresponding labels, we need to load the actual image given its filename. For this we use the map function of the Dataset class. This function applies a function, in our case the parse_func function, to every sample in a batch. The parse_func function then reads the actual image given its name, decodes and normalizes it. It also makes sure that we use the correct dimensions and data types. With the build_dataset function we can then build a Dataset obejct, which first shuffles then applies the parse_func function and finnaly batches the data. In order to iterate many times over the data set, we repeat this process indefenetly.\n",
    "\n",
    "In order to train the model, we again use a function which does one training step. This function is similar to the previous part of this exercise with the addition of the @tf.function decorator, which tells Tensorflow to build this function as a graph in order to speed up repeated calls to it. We also use the mean absolute error (MAE) as a loss for training the model. This loss can be derived in the same way as in the previous part if we assume laplacian distributed noise in out signal model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "yjMgN61ijYD2"
   },
   "outputs": [],
   "source": [
    "# Define loss function\n",
    "def loss(y, y_pred):\n",
    "    return tf.reduce_mean(tf.abs(y-y_pred))\n",
    "\n",
    "# Define training step as a complete graph\n",
    "@tf.function\n",
    "def train_step(model, optimizer, x, y):\n",
    "    with tf.GradientTape() as tape:\n",
    "        tape.watch(model.trainable_variables)\n",
    "        y_pred = model(x)\n",
    "        loss_val = loss(y, y_pred)\n",
    "    grads = tape.gradient(loss_val, model.trainable_variables)\n",
    "    optimizer.apply_gradients(zip(grads, model.trainable_variables))\n",
    "    return loss_val"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "ZuoE9caslOI1"
   },
   "source": [
    "Now we just need to instantiate an instance of our model and an optimizer in order to start training. We also build the model in order to be able to print it's summary."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 421
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 81076,
     "status": "ok",
     "timestamp": 1562140806189,
     "user": {
      "displayName": "Felix Wiewel",
      "photoUrl": "https://lh3.googleusercontent.com/-XBgpqQQzeMU/AAAAAAAAAAI/AAAAAAAAAZY/12qzl-EFXfM/s64/photo.jpg",
      "userId": "11509421729312238102"
     },
     "user_tz": -120
    },
    "id": "uQ1C4NEklURm",
    "outputId": "0e9dbade-4bcb-49c4-a245-f9b7515006cc"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"my_model\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              multiple                  608       \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            multiple                  3216      \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            multiple                  0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            multiple                  0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                multiple                  295040    \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              multiple                  8256      \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              multiple                  65        \n",
      "=================================================================\n",
      "Total params: 307,185\n",
      "Trainable params: 307,185\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "mdl = MyModel()\n",
    "opt = tf.optimizers.RMSprop(learning_rate)\n",
    "mdl.build((batch_size, 200, 200, 3))\n",
    "mdl.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "NSn4sAMflaKI"
   },
   "source": [
    "We use a similar loop as in the previous part of this exercise for training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 384
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 401389,
     "status": "ok",
     "timestamp": 1562141126510,
     "user": {
      "displayName": "Felix Wiewel",
      "photoUrl": "https://lh3.googleusercontent.com/-XBgpqQQzeMU/AAAAAAAAAAI/AAAAAAAAAZY/12qzl-EFXfM/s64/photo.jpg",
      "userId": "11509421729312238102"
     },
     "user_tz": -120
    },
    "id": "C1HPWl5Cl0DR",
    "outputId": "a7d42834-6847-4872-9dac-6170da458b06"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 1 Training loss: 16.164 Validation loss 11.932\n",
      "Epoch: 2 Training loss: 13.469 Validation loss 11.224\n",
      "Epoch: 3 Training loss: 11.63 Validation loss 9.3953\n",
      "Epoch: 4 Training loss: 10.516 Validation loss 10.646\n",
      "Epoch: 5 Training loss: 9.7195 Validation loss 8.7859\n",
      "Epoch: 6 Training loss: 9.2718 Validation loss 9.1778\n",
      "Epoch: 7 Training loss: 8.8482 Validation loss 8.3116\n",
      "Epoch: 8 Training loss: 8.5844 Validation loss 7.5919\n",
      "Epoch: 9 Training loss: 8.3456 Validation loss 7.5368\n",
      "Epoch: 10 Training loss: 8.0617 Validation loss 8.7846\n",
      "Epoch: 11 Training loss: 7.9239 Validation loss 6.7009\n",
      "Epoch: 12 Training loss: 7.7069 Validation loss 6.6426\n",
      "Epoch: 13 Training loss: 7.485 Validation loss 7.3148\n",
      "Epoch: 14 Training loss: 7.3211 Validation loss 6.5898\n",
      "Epoch: 15 Training loss: 7.107 Validation loss 6.3446\n",
      "Epoch: 16 Training loss: 6.9443 Validation loss 6.7848\n",
      "Epoch: 17 Training loss: 6.7349 Validation loss 6.4691\n",
      "Epoch: 18 Training loss: 6.6226 Validation loss 6.315\n",
      "Epoch: 19 Training loss: 6.4756 Validation loss 8.6024\n",
      "Epoch: 20 Training loss: 6.3307 Validation loss 6.5534\n"
     ]
    }
   ],
   "source": [
    "# Run training\n",
    "epoch = 0\n",
    "train_loss = 0.0\n",
    "train_iters = 0\n",
    "for train_images, train_labels in train_ds:\n",
    "    train_loss += train_step(mdl, opt, train_images, train_labels)\n",
    "    train_iters += 1\n",
    "    if train_iters == int(N_training_examples/batch_size):\n",
    "        epoch += 1\n",
    "        val_loss = 0.0\n",
    "        val_iters = 0\n",
    "        for val_images, val_labels in validation_ds:\n",
    "            y_pred = mdl(val_images)\n",
    "            loss_val = loss(val_labels, y_pred)\n",
    "            val_loss += loss_val\n",
    "            val_iters += 1\n",
    "            if val_iters == int(N_validation_examples/batch_size):\n",
    "                print(\"Epoch: {} Training loss: {:.5} Validation loss {:.5}\"\n",
    "                      .format(epoch, train_loss/train_iters, val_loss/val_iters))\n",
    "                break\n",
    "        train_loss = 0.0\n",
    "        train_iters = 0\n",
    "    if epoch == N_epochs:\n",
    "        break"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "IkdWtN69Khno"
   },
   "source": [
    "Testing the model on the test data set is also similar to the previous part of this exercise."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 420291,
     "status": "ok",
     "timestamp": 1562141145422,
     "user": {
      "displayName": "Felix Wiewel",
      "photoUrl": "https://lh3.googleusercontent.com/-XBgpqQQzeMU/AAAAAAAAAAI/AAAAAAAAAZY/12qzl-EFXfM/s64/photo.jpg",
      "userId": "11509421729312238102"
     },
     "user_tz": -120
    },
    "id": "t13d1a0xKrx3",
    "outputId": "5f59c6e6-2b6f-4422-9f70-80883544bf60"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 7.7438\n"
     ]
    }
   ],
   "source": [
    "# Test model\n",
    "Num_test_batches = int((len(files)-N_validation_examples-N_validation_examples)/batch_size)\n",
    "test_loss = 0.0\n",
    "test_iters = 0\n",
    "for test_images, test_labels in test_ds:\n",
    "    y_pred = mdl(test_images)\n",
    "    test_loss += loss(test_labels, y_pred)\n",
    "    test_iters += 1\n",
    "    if test_iters == Num_test_batches:\n",
    "        print(\"Test loss: {:.5}\".format(test_loss/test_iters))\n",
    "        break"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "rMOKzIG8SUbq"
   },
   "source": [
    "We again can compare the training, validation and test losses in order to verify if our model is overfitting or not. We can now also upload our own images and predict on them. For this we need to upload an image which contains a quadratic potrait with the face aligned in the center."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 85,
     "resources": {
      "http://localhost:8080/nbextensions/google.colab/files.js": {
       "data": "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",
       "headers": [
        [
         "content-type",
         "application/javascript"
        ]
       ],
       "ok": true,
       "status": 200,
       "status_text": "OK"
      }
     }
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 4597,
     "status": "ok",
     "timestamp": 1562144663653,
     "user": {
      "displayName": "Felix Wiewel",
      "photoUrl": "https://lh3.googleusercontent.com/-XBgpqQQzeMU/AAAAAAAAAAI/AAAAAAAAAZY/12qzl-EFXfM/s64/photo.jpg",
      "userId": "11509421729312238102"
     },
     "user_tz": -120
    },
    "id": "hmZMBW8eSUI9",
    "outputId": "a20b644c-cada-465a-d0c2-76dae20bade3"
   },
   "outputs": [
    {
     "ename": "ModuleNotFoundError",
     "evalue": "No module named 'google.colab'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-12-d90c8cda1d76>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0mgoogle\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcolab\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mfiles\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      2\u001b[0m \u001b[0muploaded\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfiles\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mupload\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0mfn\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlist\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0muploaded\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mkeys\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'google.colab'"
     ]
    }
   ],
   "source": [
    "from google.colab import files\n",
    "uploaded = files.upload()\n",
    "fn = list(uploaded.keys())[0]\n",
    "  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "0qHY4pl1Suzv"
   },
   "source": [
    "Now we do a prediction  and plot the result."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 281
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 1089,
     "status": "ok",
     "timestamp": 1562144668137,
     "user": {
      "displayName": "Felix Wiewel",
      "photoUrl": "https://lh3.googleusercontent.com/-XBgpqQQzeMU/AAAAAAAAAAI/AAAAAAAAAZY/12qzl-EFXfM/s64/photo.jpg",
      "userId": "11509421729312238102"
     },
     "user_tz": -120
    },
    "id": "GW1trSa7SxfF",
    "outputId": "e0de8945-d38c-4783-8258-066de51d1ff3"
   },
   "outputs": [],
   "source": [
    "# Load and predict on an image\n",
    "image_string = tf.io.read_file(fn)\n",
    "image_decoded = tf.expand_dims(tf.io.decode_jpeg(image_string), axis=0)\n",
    "image = tf.image.resize(tf.cast(image_decoded, tf.float32)/255.0, [200, 200])\n",
    "age = mdl(image)\n",
    "\n",
    "# Plot image and prediction\n",
    "plt.imshow(np.squeeze(image.numpy()))\n",
    "plt.title(\"Age prediction: {:.3}\".format(np.squeeze(age.numpy())))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "uULbhmkTIK4y"
   },
   "source": [
    "## Conclusion\n",
    "In this part of the exercise we have applied the concept of regression to the problem of estimating the age of a person from his or her potrait. In this process we have seen how to load images from their filenames with the map function of the Dataset class. With this, images are loaded just in time as they are needed, which is beneficial if you are working with a large number of high resolution images that do not fit into your system memory all at once. We also used Keras in order to build our model from predefined layers. If the layers you need for building your model are available through Keras, it is highly recommended to use those in order to speed up develpment time and reduce possible error sources. We also observed that the labeling of the individual images is not optimal for our task, since it is rounded to integer ages and in the setting of regression we would rather like to have continuous age labels."
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "Age_regression_solution.ipynb",
   "provenance": [
    {
     "file_id": "1QGa9t8a7AXIT8j0MUNIreJBZEEVgEGO-",
     "timestamp": 1559828019631
    },
    {
     "file_id": "1jurik7CuCW9wO7cQKGBYj2Dfz-nJkjBc",
     "timestamp": 1559819843632
    }
   ],
   "version": "0.3.2"
  },
  "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": "3.8.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}