Siamese-Neural-Network-MNIST.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Siamese Neural Network with Triplet Loss trained on MNIST\n",
    "## Cameron Trotter\n",
    "### c.trotter2@ncl.ac.uk\n",
    "\n",
    "This notebook builds an SNN to determine similarity scores between MNIST digits using a triplet loss function. The use of class prototypes at inference time is also explored. \n",
    "\n",
    "This notebook is based heavily on the approach described in [this Coursera course](https://www.coursera.org/learn/siamese-network-triplet-loss-keras/), which in turn is based on the [FaceNet](https://arxiv.org/abs/1503.03832) paper. Any uses of open-source code are linked throughout where utilised. \n",
    "\n",
    "For an in-depth guide to understand this code, and the theory behind it, please see LINK.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TF 1.14 gives lots of warnings for deprecations ready for the switch to TF 2.0\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import random\n",
    "import os\n",
    "import glob\n",
    "\n",
    "from datetime import datetime\n",
    "from tensorflow.keras.models import model_from_json\n",
    "from tensorflow.keras.callbacks import Callback, CSVLogger, ModelCheckpoint\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from tensorflow.keras.layers import Activation, Input, concatenate\n",
    "from tensorflow.keras.layers import Layer, BatchNormalization, MaxPooling2D, Concatenate, Lambda, Flatten, Dense\n",
    "from tensorflow.keras.initializers import glorot_uniform, he_uniform\n",
    "from tensorflow.keras.regularizers import l2\n",
    "from tensorflow.keras.utils import multi_gpu_model\n",
    "\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.metrics import roc_curve, roc_auc_score\n",
    "import math\n",
    "from pylab import dist\n",
    "import json\n",
    "\n",
    "from tensorflow.python.client import device_lib\n",
    "import matplotlib.gridspec as gridspec"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import the data and reshape for use with the SNN\n",
    "\n",
    "The data loaded in must be in the same format as `tf.keras.datasets.mnist.load_data()`, that is `(x_train, y_train), (x_test, y_test)`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()\n",
    "\n",
    "num_classes = len(np.unique(y_train))\n",
    "\n",
    "x_train_w = x_train.shape[1] # (60000, 28, 28)\n",
    "x_train_h = x_train.shape[2]\n",
    "x_test_w = x_test.shape[1]\n",
    "x_test_h = x_test.shape[2]\n",
    "\n",
    "x_train_w_h = x_train_w * x_train_h # 28 * 28 = 784\n",
    "x_test_w_h = x_test_w * x_test_h\n",
    "\n",
    "x_train = np.reshape(x_train, (x_train.shape[0], x_train_w_h))/255. # (60000, 784)\n",
    "x_test = np.reshape(x_test, (x_test.shape[0], x_test_w_h))/255."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Plotting the triplets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x144 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_triplets(examples):\n",
    "    plt.figure(figsize=(6, 2))\n",
    "    for i in range(3):\n",
    "        plt.subplot(1, 3, 1 + i)\n",
    "        plt.imshow(np.reshape(examples[i], (x_train_w, x_train_h)), cmap='binary')\n",
    "        plt.xticks([])\n",
    "        plt.yticks([])\n",
    "    plt.show()\n",
    "    \n",
    "plot_triplets([x_train[0], x_train[1], x_train[2]])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  Create triplet batches\n",
    "\n",
    "Random batches are generated by `create_batch`. Semi-hard triplet batches are generated by `create_batch_hard`.\n",
    "\n",
    "Semi-Hard: dist(A, P) < dist(A, N) < dist(A, P) + margin. Using only easy triplets will lead to no learning. Hard triplets generate high loss and have high impact on training parameters, but may cause any mislabelled data to cause too much of a weight change. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_batch(batch_size=256, split = \"train\"):\n",
    "    x_anchors = np.zeros((batch_size, x_train_w_h))\n",
    "    x_positives = np.zeros((batch_size, x_train_w_h))\n",
    "    x_negatives = np.zeros((batch_size, x_train_w_h))\n",
    "    \n",
    "    if split ==\"train\":\n",
    "        data = x_train\n",
    "        data_y = y_train\n",
    "    else:\n",
    "        data = x_test\n",
    "        data_y = y_test\n",
    "    \n",
    "    for i in range(0, batch_size):\n",
    "        # We need to find an anchor, a positive example and a negative example\n",
    "        random_index = random.randint(0, data.shape[0] - 1)\n",
    "        x_anchor = data[random_index]\n",
    "        y = data_y[random_index]\n",
    "        \n",
    "        indices_for_pos = np.squeeze(np.where(data_y == y))\n",
    "        indices_for_neg = np.squeeze(np.where(data_y != y))\n",
    "        \n",
    "        x_positive = data[indices_for_pos[random.randint(0, len(indices_for_pos) - 1)]]\n",
    "        x_negative = data[indices_for_neg[random.randint(0, len(indices_for_neg) - 1)]]\n",
    "        \n",
    "        x_anchors[i] = x_anchor\n",
    "        x_positives[i] = x_positive\n",
    "        x_negatives[i] = x_negative\n",
    "        \n",
    "    return [x_anchors, x_positives, x_negatives]\n",
    "\n",
    "\n",
    "def create_hard_batch(batch_size, num_hard, split = \"train\"):\n",
    "    \n",
    "    x_anchors = np.zeros((batch_size, x_train_w_h))\n",
    "    x_positives = np.zeros((batch_size, x_train_w_h))\n",
    "    x_negatives = np.zeros((batch_size, x_train_w_h))\n",
    "    \n",
    "    if split ==\"train\":\n",
    "        data = x_train\n",
    "        data_y = y_train\n",
    "    else:\n",
    "        data = x_test\n",
    "        data_y = y_test\n",
    "    \n",
    "    # Generate num_hard number of hard examples:\n",
    "    hard_batches = [] \n",
    "    batch_losses = []\n",
    "    \n",
    "    rand_batches = []\n",
    "    \n",
    "    # Get some random batches\n",
    "    for i in range(0, batch_size):\n",
    "        hard_batches.append(create_batch(1, split))\n",
    "        \n",
    "        A_emb = embedding_model.predict(hard_batches[i][0])\n",
    "        P_emb = embedding_model.predict(hard_batches[i][1])\n",
    "        N_emb = embedding_model.predict(hard_batches[i][2])\n",
    "        \n",
    "        # Compute d(A, P) - d(A, N) for each selected batch\n",
    "        batch_losses.append(np.sum(np.square(A_emb-P_emb),axis=1) - np.sum(np.square(A_emb-N_emb),axis=1))\n",
    "    \n",
    "    # Sort batch_loss by distance, highest first, and keep num_hard of them\n",
    "    hard_batch_selections = [x for _, x in sorted(zip(batch_losses,hard_batches), key=lambda x: x[0])]\n",
    "    hard_batches = hard_batch_selections[:num_hard]\n",
    "    \n",
    "    # Get batch_size - num_hard number of random examples\n",
    "    num_rand = batch_size - num_hard\n",
    "    for i in range(0, num_rand):\n",
    "        rand_batch = create_batch(1, split)\n",
    "        rand_batches.append(rand_batch)\n",
    "    \n",
    "    selections = hard_batches + rand_batches\n",
    "    \n",
    "    for i in range(0, len(selections)):\n",
    "        x_anchors[i] = selections[i][0]\n",
    "        x_positives[i] = selections[i][1]\n",
    "        x_negatives[i] = selections[i][2]\n",
    "        \n",
    "    return [x_anchors, x_positives, x_negatives]\n",
    "    \n",
    "    "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create the Embedding Model\n",
    "\n",
    "This model takes in input image and generates some `emb_size`-dimensional embedding for the image, plotted on some latent space.\n",
    "\n",
    "The untrained model's embedding space is stored for later use when comparing clustering between the untrained and the trained model using PCA, based on [this notebook](https://github.com/AdrianUng/keras-triplet-loss-mnist/blob/master/Triplet_loss_KERAS_semi_hard_from_TF.ipynb)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_embedding_model(emb_size):\n",
    "    embedding_model = tf.keras.models.Sequential([\n",
    "        Dense(4096,\n",
    "              activation='relu',\n",
    "              kernel_regularizer=l2(1e-3),\n",
    "              kernel_initializer='he_uniform',\n",
    "              input_shape=(x_train_w_h,)),\n",
    "        Dense(emb_size,\n",
    "              activation=None,\n",
    "              kernel_regularizer=l2(1e-3),\n",
    "              kernel_initializer='he_uniform')\n",
    "    ])\n",
    "\n",
    "    embedding_model.summary()\n",
    "    \n",
    "    return embedding_model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create the SNN\n",
    "\n",
    "This model takes a triplet image input, passes them to the embedding model for embedding, then concats them together for the loss function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_SNN(embedding_model):\n",
    "\n",
    "    input_anchor = tf.keras.layers.Input(shape=(x_train_w_h,))\n",
    "    input_positive = tf.keras.layers.Input(shape=(x_train_w_h,))\n",
    "    input_negative = tf.keras.layers.Input(shape=(x_train_w_h,))\n",
    "\n",
    "    embedding_anchor = embedding_model(input_anchor)\n",
    "    embedding_positive = embedding_model(input_positive)\n",
    "    embedding_negative = embedding_model(input_negative)\n",
    "\n",
    "    output = tf.keras.layers.concatenate([embedding_anchor, embedding_positive, \n",
    "                                          embedding_negative], axis=1)\n",
    "\n",
    "    siamese_net = tf.keras.models.Model([input_anchor, input_positive, input_negative], \n",
    "                                        output)\n",
    "    siamese_net.summary()\n",
    "    \n",
    "    return siamese_net"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create the Triplet Loss Function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "def triplet_loss(y_true, y_pred):\n",
    "    anchor, positive, negative = y_pred[:,:emb_size], y_pred[:,emb_size:2*emb_size],y_pred[:,2*emb_size:]\n",
    "    positive_dist = tf.reduce_mean(tf.square(anchor - positive), axis=1)\n",
    "    negative_dist = tf.reduce_mean(tf.square(anchor - negative), axis=1)\n",
    "    return tf.maximum(positive_dist - negative_dist + alpha, 0.)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data Generator\n",
    "\n",
    "This function creates hard batches for the network to train on. `y` is required by TF but not by our model, so just return a filler to keep TF happy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "def data_generator(batch_size=256, num_hard=50, split=\"train\"):\n",
    "    while True:\n",
    "        x = create_hard_batch(batch_size, num_hard, split)\n",
    "        y = np.zeros((batch_size, 3*emb_size))\n",
    "        yield x, y"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluation\n",
    "\n",
    "Allows for the model's metrics to be visualised and evaluated. Based on [this Medium post](https://medium.com/@crimy/one-shot-learning-siamese-networks-and-triplet-loss-with-keras-2885ed022352) and [this GitHub notebook](https://github.com/asagar60/One-Shot-Learning/blob/master/Omniglot_data/One_shot_implementation.ipynb)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "def compute_dist(a,b):\n",
    "    return np.linalg.norm(a-b)\n",
    "\n",
    "def compute_probs(network,X,Y):\n",
    "    '''\n",
    "    Input\n",
    "        network : current NN to compute embeddings\n",
    "        X : tensor of shape (m,w,h,1) containing pics to evaluate\n",
    "        Y : tensor of shape (m,) containing true class\n",
    "        \n",
    "    Returns\n",
    "        probs : array of shape (m,m) containing distances\n",
    "    \n",
    "    '''\n",
    "    m = X.shape[0]\n",
    "    nbevaluation = int(m*(m-1)/2)\n",
    "    probs = np.zeros((nbevaluation))\n",
    "    y = np.zeros((nbevaluation))\n",
    "    \n",
    "    #Compute all embeddings for all imgs with current embedding network\n",
    "    embeddings = embedding_model.predict(X)\n",
    "    \n",
    "    k = 0\n",
    "    \n",
    "    # For each img in the evaluation set\n",
    "    for i in range(m):\n",
    "            # Against all other images\n",
    "            for j in range(i+1,m):\n",
    "                # compute the probability of being the right decision : it should be 1 for right class, 0 for all other classes\n",
    "                probs[k] = -compute_dist(embeddings[i,:],embeddings[j,:])\n",
    "                if (Y[i]==Y[j]):\n",
    "                    y[k] = 1\n",
    "                    #print(\"{3}:{0} vs {1} : \\t\\t\\t{2}\\tSAME\".format(i,j,probs[k],k, Y[i], Y[j]))\n",
    "                else:\n",
    "                    y[k] = 0\n",
    "                    #print(\"{3}:{0} vs {1} : {2}\\tDIFF\".format(i,j,probs[k],k, Y[i], Y[j]))\n",
    "                k += 1\n",
    "    return probs, y\n",
    "\n",
    "\n",
    "def compute_metrics(probs,yprobs):\n",
    "    '''\n",
    "    Returns\n",
    "        fpr : Increasing false positive rates such that element i is the false positive rate of predictions with score >= thresholds[i]\n",
    "        tpr : Increasing true positive rates such that element i is the true positive rate of predictions with score >= thresholds[i].\n",
    "        thresholds : Decreasing thresholds on the decision function used to compute fpr and tpr. thresholds[0] represents no instances being predicted and is arbitrarily set to max(y_score) + 1\n",
    "        auc : Area Under the ROC Curve metric\n",
    "    '''\n",
    "    # calculate AUC\n",
    "    auc = roc_auc_score(yprobs, probs)\n",
    "    # calculate roc curve\n",
    "    fpr, tpr, thresholds = roc_curve(yprobs, probs)\n",
    "    \n",
    "    return fpr, tpr, thresholds,auc\n",
    "\n",
    "def draw_roc(fpr, tpr,thresholds, auc):\n",
    "    #find threshold\n",
    "    targetfpr=1e-3\n",
    "    _, idx = find_nearest(fpr,targetfpr)\n",
    "    threshold = thresholds[idx]\n",
    "    recall = tpr[idx]\n",
    "    \n",
    "    \n",
    "    # plot no skill\n",
    "    plt.plot([0, 1], [0, 1], linestyle='--')\n",
    "    # plot the roc curve for the model\n",
    "    plt.plot(fpr, tpr, marker='.')\n",
    "    plt.title('AUC: {0:.3f}\\nSensitivity : {2:.1%} @FPR={1:.0e}\\nThreshold={3})'.format(auc,targetfpr,recall,abs(threshold) ))\n",
    "    # show the plot\n",
    "    plt.show()\n",
    "    \n",
    "def find_nearest(array,value):\n",
    "    idx = np.searchsorted(array, value, side=\"left\")\n",
    "    if idx > 0 and (idx == len(array) or math.fabs(value - array[idx-1]) < math.fabs(value - array[idx])):\n",
    "        return array[idx-1],idx-1\n",
    "    else:\n",
    "        return array[idx],idx\n",
    "    \n",
    "def draw_interdist(network, epochs):\n",
    "    interdist = compute_interdist(network)\n",
    "    \n",
    "    data = []\n",
    "    for i in range(num_classes):\n",
    "        data.append(np.delete(interdist[i,:],[i]))\n",
    "\n",
    "    fig, ax = plt.subplots()\n",
    "    ax.set_title('Evaluating embeddings distance from each other after {0} epochs'.format(epochs))\n",
    "    ax.set_ylim([0,3])\n",
    "    plt.xlabel('Classes')\n",
    "    plt.ylabel('Distance')\n",
    "    ax.boxplot(data,showfliers=False,showbox=True)\n",
    "    locs, labels = plt.xticks()\n",
    "    plt.xticks(locs,np.arange(num_classes))\n",
    "\n",
    "    plt.show()\n",
    "    \n",
    "    \n",
    "def compute_interdist(network):\n",
    "    '''\n",
    "    Computes sum of distances between all classes embeddings on our reference test image: \n",
    "        d(0,1) + d(0,2) + ... + d(0,9) + d(1,2) + d(1,3) + ... d(8,9)\n",
    "        A good model should have a large distance between all theses embeddings\n",
    "        \n",
    "    Returns:\n",
    "        array of shape (num_classes,num_classes) \n",
    "    '''\n",
    "    res = np.zeros((num_classes,num_classes))\n",
    "    \n",
    "    ref_images = np.zeros((num_classes, x_test_w_h))\n",
    "    \n",
    "    #generates embeddings for reference images\n",
    "    for i in range(num_classes):\n",
    "        ref_images[i,:] = x_test[i]\n",
    "    \n",
    "    ref_embeddings = network.predict(ref_images)\n",
    "    \n",
    "    for i in range(num_classes):\n",
    "        for j in range(num_classes):\n",
    "            res[i,j] = dist(ref_embeddings[i],ref_embeddings[j])\n",
    "    return res\n",
    "\n",
    "def DrawTestImage(network, images, refidx=0):\n",
    "    '''\n",
    "    Evaluate some pictures vs some samples in the test set\n",
    "        image must be of shape(1,w,h,c)\n",
    "    \n",
    "    Returns\n",
    "        scores : result of the similarity scores with the basic images => (N)\n",
    "    \n",
    "    '''\n",
    "    nbimages = images.shape[0]\n",
    "    \n",
    "    \n",
    "    #generates embedings for given images\n",
    "    image_embedings = network.predict(images)\n",
    "    \n",
    "    #generates embedings for reference images\n",
    "    ref_images = np.zeros((num_classes,x_test_w_h))\n",
    "    for i in range(num_classes):\n",
    "        images_at_this_index_are_of_class_i = np.squeeze(np.where(y_test == i))\n",
    "        ref_images[i,:] = x_test[images_at_this_index_are_of_class_i[refidx]]\n",
    "        \n",
    "        \n",
    "    ref_embedings = network.predict(ref_images)\n",
    "            \n",
    "    for i in range(nbimages):\n",
    "        # Prepare the figure\n",
    "        fig=plt.figure(figsize=(16,2))\n",
    "        subplot = fig.add_subplot(1,num_classes+1,1)\n",
    "        plt.axis(\"off\")\n",
    "        plotidx = 2\n",
    "            \n",
    "        # Draw this image    \n",
    "        plt.imshow(np.reshape(images[i], (x_train_w, x_train_h)),vmin=0, vmax=1,cmap='Greys')\n",
    "        subplot.title.set_text(\"Test image\")\n",
    "            \n",
    "        for ref in range(num_classes):\n",
    "            #Compute distance between this images and references\n",
    "            dist = compute_dist(image_embedings[i,:],ref_embedings[ref,:])\n",
    "            #Draw\n",
    "            subplot = fig.add_subplot(1,num_classes+1,plotidx)\n",
    "            plt.axis(\"off\")\n",
    "            plt.imshow(np.reshape(ref_images[ref, :], (x_train_w, x_train_h)),vmin=0, vmax=1,cmap='Greys')\n",
    "            subplot.title.set_text((\"Class {0}\\n{1:.3e}\".format(ref,dist)))\n",
    "            plotidx += 1\n",
    "\n",
    "def generate_prototypes(x_data, y_data, embedding_model):\n",
    "    classes = np.unique(y_data)\n",
    "    prototypes = {}\n",
    "\n",
    "    for c in classes:\n",
    "        #c = classes[0]\n",
    "        # Find all images of the chosen test class\n",
    "        locations_of_c = np.where(y_data == c)[0]\n",
    "\n",
    "        imgs_of_c = x_data[locations_of_c]\n",
    "\n",
    "        imgs_of_c_embeddings = embedding_model.predict(imgs_of_c)\n",
    "\n",
    "        # Get the median of the embeddings to generate a prototype for the class (reshaping for PCA)\n",
    "        prototype_for_c = np.median(imgs_of_c_embeddings, axis = 0).reshape(1, -1)\n",
    "        # Add it to the prototype dict\n",
    "        prototypes[c] = prototype_for_c\n",
    "        \n",
    "    return prototypes\n",
    "         \n",
    "def test_one_shot_prototypes(network, sample_embeddings):\n",
    "    distances_from_img_to_test_against = []\n",
    "    # As the img to test against is in index 0, we compare distances between img@0 and all others\n",
    "    for i in range(1, len(sample_embeddings)):\n",
    "        distances_from_img_to_test_against.append(compute_dist(sample_embeddings[0], sample_embeddings[i]))\n",
    "    # As the correct img will be at distances_from_img_to_test_against index 0 (sample_imgs index 1),\n",
    "    # If the smallest distance in distances_from_img_to_test_against is at index 0, \n",
    "    # we know the one shot test got the right answer\n",
    "    is_min = distances_from_img_to_test_against[0] == min(distances_from_img_to_test_against)\n",
    "    is_max = distances_from_img_to_test_against[0] == max(distances_from_img_to_test_against)\n",
    "    return int(is_min and not is_max)\n",
    "    \n",
    "def n_way_accuracy_prototypes(n_val, n_way, network):\n",
    "    num_correct = 0\n",
    "    \n",
    "    for val_step in range(n_val):\n",
    "        num_correct += load_one_shot_test_batch_prototypes(n_way, network)\n",
    "        \n",
    "    accuracy = num_correct / n_val * 100\n",
    "        \n",
    "    return accuracy\n",
    "\n",
    "def load_one_shot_test_batch_prototypes(n_way, network):\n",
    "    \n",
    "    labels = np.unique(y_test)\n",
    "    # Reduce the label set down from size n_classes to n_samples \n",
    "    labels = np.random.choice(labels, size = n_way, replace = False)\n",
    "\n",
    "    # Choose a class as the test image\n",
    "    label = random.choice(labels)\n",
    "    # Find all images of the chosen test class\n",
    "    imgs_of_label = np.where(y_test == label)[0]\n",
    "\n",
    "    # Randomly select a test image of the selected class, return it's index\n",
    "    img_of_label_idx = random.choice(imgs_of_label)\n",
    "\n",
    "    # Expand the array at the selected indexes into useable images\n",
    "    img_of_label = np.expand_dims(x_test[img_of_label_idx],axis=0)\n",
    "    \n",
    "    sample_embeddings = []\n",
    "    # Get the anchor image embedding\n",
    "    anchor_prototype = network.predict(img_of_label)\n",
    "    sample_embeddings.append(anchor_prototype)\n",
    "    \n",
    "    # Get the prototype embedding for the positive class\n",
    "    positive_prototype = prototypes[label]\n",
    " \n",
    "    sample_embeddings.append(positive_prototype)\n",
    "    \n",
    "    # Get the negative prototype embeddings\n",
    "    # Remove the selected test class from the list of labels based on it's index \n",
    "    label_idx_in_labels = np.where(labels == label)[0]\n",
    "    other_labels = np.delete(labels, label_idx_in_labels)\n",
    "    \n",
    "    # Get the embedding for each of the remaining negatives\n",
    "    for other_label in other_labels:\n",
    "        negative_prototype = prototypes[other_label]\n",
    "        sample_embeddings.append(negative_prototype)\n",
    "                \n",
    "    correct = test_one_shot_prototypes(network, sample_embeddings)\n",
    "\n",
    "    return correct\n",
    "\n",
    "\n",
    "def visualise_n_way_prototypes(n_samples, network):\n",
    "    labels = np.unique(y_test)\n",
    "    # Reduce the label set down from size n_classes to n_samples \n",
    "    labels = np.random.choice(labels, size = n_samples, replace = False)\n",
    "\n",
    "    # Choose a class as the test image\n",
    "    label = random.choice(labels)\n",
    "    # Find all images of the chosen test class\n",
    "    imgs_of_label = np.where(y_test == label)[0]\n",
    "\n",
    "    # Randomly select a test image of the selected class, return it's index\n",
    "    img_of_label_idx = random.choice(imgs_of_label)\n",
    "\n",
    "    # Get another image idx that we know is of the test class for the sample set\n",
    "    label_sample_img_idx = random.choice(imgs_of_label)\n",
    "\n",
    "    # Expand the array at the selected indexes into useable images\n",
    "    img_of_label = np.expand_dims(x_test[img_of_label_idx],axis=0)\n",
    "    label_sample_img = np.expand_dims(x_test[label_sample_img_idx],axis=0)\n",
    "    \n",
    "    # Make the first img in the sample set the chosen test image, the second the other image\n",
    "    sample_imgs = np.empty((0, x_test_w_h))\n",
    "    sample_imgs = np.append(sample_imgs, img_of_label, axis=0)\n",
    "    sample_imgs = np.append(sample_imgs, label_sample_img, axis=0)\n",
    "    \n",
    "    sample_embeddings = []\n",
    "    \n",
    "    # Get the anchor embedding image\n",
    "    anchor_prototype = network.predict(img_of_label)\n",
    "    sample_embeddings.append(anchor_prototype)\n",
    "    \n",
    "    # Get the prototype embedding for the positive class\n",
    "    positive_prototype = prototypes[label]\n",
    "    sample_embeddings.append(positive_prototype)\n",
    "\n",
    "    # Get the negative prototype embeddings\n",
    "    # Remove the selected test class from the list of labels based on it's index \n",
    "    label_idx_in_labels = np.where(labels == label)[0]\n",
    "    other_labels = np.delete(labels, label_idx_in_labels)\n",
    "    # Get the embedding for each of the remaining negatives\n",
    "    for other_label in other_labels:\n",
    "        negative_prototype = prototypes[other_label]\n",
    "        sample_embeddings.append(negative_prototype)\n",
    "        \n",
    "        # Find all images of the other class\n",
    "        imgs_of_other_label = np.where(y_test == other_label)[0]\n",
    "        # Randomly select an image of the selected class, return it's index\n",
    "        another_sample_img_idx = random.choice(imgs_of_other_label)\n",
    "        # Expand the array at the selected index into useable images\n",
    "        another_sample_img = np.expand_dims(x_test[another_sample_img_idx],axis=0)\n",
    "        # Add the image to the support set\n",
    "        sample_imgs = np.append(sample_imgs, another_sample_img, axis=0)\n",
    "    \n",
    "    distances_from_img_to_test_against = []\n",
    "    \n",
    "    # As the img to test against is in index 0, we compare distances between img@0 and all others\n",
    "    for i in range(1, len(sample_embeddings)):\n",
    "        distances_from_img_to_test_against.append(compute_dist(sample_embeddings[0], sample_embeddings[i]))\n",
    "        \n",
    "    # + 1 as distances_from_img_to_test_against doesn't include the test image\n",
    "    min_index = distances_from_img_to_test_against.index(min(distances_from_img_to_test_against)) + 1\n",
    "    \n",
    "    return sample_imgs, min_index\n",
    "\n",
    "def evaluate(embedding_model, epochs = 0):\n",
    "    probs,yprob = compute_probs(embedding_model, x_test[:500, :], y_test[:500])\n",
    "    fpr, tpr, thresholds, auc = compute_metrics(probs,yprob)\n",
    "    draw_roc(fpr, tpr, thresholds, auc)\n",
    "    draw_interdist(embedding_model, epochs)\n",
    "\n",
    "    for i in range(3):\n",
    "        DrawTestImage(embedding_model, np.expand_dims(x_train[i],axis=0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model Training Setup\n",
    "\n",
    "FaceNet, the original triplet batch paper, draws a large random sample of triplets respecting the class distribution then picks N/2 hard and N/2 random samples (N = batch size), along with an `alpha` of 0.2\n",
    "\n",
    "Logs out to Tensorboard, callback adapted from https://stackoverflow.com/a/52581175.\n",
    "\n",
    "Saves best model only based on a validation loss. Adapted from https://stackoverflow.com/a/58103272."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Generating embedding model... \n",
      "\n",
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_4 (Dense)              (None, 4096)              3215360   \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 10)                40970     \n",
      "=================================================================\n",
      "Total params: 3,256,330\n",
      "Trainable params: 3,256,330\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "\n",
      "Generating SNN... \n",
      "\n",
      "Model: \"model_4\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_7 (InputLayer)            [(None, 784)]        0                                            \n",
      "__________________________________________________________________________________________________\n",
      "input_8 (InputLayer)            [(None, 784)]        0                                            \n",
      "__________________________________________________________________________________________________\n",
      "input_9 (InputLayer)            [(None, 784)]        0                                            \n",
      "__________________________________________________________________________________________________\n",
      "sequential_2 (Sequential)       (None, 10)           3256330     input_7[0][0]                    \n",
      "                                                                 input_8[0][0]                    \n",
      "                                                                 input_9[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_2 (Concatenate)     (None, 30)           0           sequential_2[1][0]               \n",
      "                                                                 sequential_2[2][0]               \n",
      "                                                                 sequential_2[3][0]               \n",
      "==================================================================================================\n",
      "Total params: 3,256,330\n",
      "Trainable params: 3,256,330\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n",
      "\n",
      "Evaluating the model without training for a baseline...\n",
      "\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"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x144 with 11 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x144 with 11 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x144 with 11 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Hyperparams\n",
    "batch_size = 256\n",
    "epochs = 100\n",
    "steps_per_epoch = int(x_train.shape[0]/batch_size)\n",
    "val_steps = int(x_test.shape[0]/batch_size)\n",
    "alpha = 0.2\n",
    "num_hard = int(batch_size * 0.5) # Number of semi-hard triplet examples in the batch\n",
    "lr = 0.00006\n",
    "optimiser = 'Adam'\n",
    "emb_size = 10\n",
    "\n",
    "with tf.device(\"/cpu:0\"):\n",
    "    # Create the embedding model\n",
    "    print(\"Generating embedding model... \\n\")\n",
    "    embedding_model = create_embedding_model(emb_size)\n",
    "    \n",
    "    print(\"\\nGenerating SNN... \\n\")\n",
    "    # Create the SNN\n",
    "    siamese_net = create_SNN(embedding_model)\n",
    "    # Compile the SNN\n",
    "    optimiser_obj = Adam(lr = lr)\n",
    "    siamese_net.compile(loss=triplet_loss, optimizer= optimiser_obj)\n",
    "    \n",
    "    # Store visualisations of the embeddings using PCA for display next to \"after training\" for comparisons\n",
    "    num_vis = 500 # Take only the first num_vis elements of the test set to visualise\n",
    "    embeddings_before_train = embedding_model.predict(x_test[:num_vis, :])\n",
    "    pca = PCA(n_components=2)\n",
    "    decomposed_embeddings_before = pca.fit_transform(embeddings_before_train)\n",
    "\n",
    "\n",
    "# Display evaluation the untrained model\n",
    "print(\"\\nEvaluating the model without training for a baseline...\\n\")\n",
    "evaluate(embedding_model)\n",
    "\n",
    "# Set up logging directory\n",
    "## Use date-time as logdir name:\n",
    "#dt = datetime.now().strftime(\"%Y%m%dT%H%M\")\n",
    "#logdir = os.path.join(\"PATH/TO/LOGDIR\",dt)\n",
    "\n",
    "## Use a custom non-dt name:\n",
    "name = \"snn-example-run\"\n",
    "logdir = os.path.join(\"PATH/TO/LOGDIR\",name)\n",
    "\n",
    "if not os.path.exists(logdir):\n",
    "    os.mkdir(logdir)\n",
    "\n",
    "## Callbacks:\n",
    "# Create the TensorBoard callback\n",
    "tensorboard = tf.keras.callbacks.TensorBoard(\n",
    "    log_dir = logdir,\n",
    "    histogram_freq=0,\n",
    "    batch_size=batch_size,\n",
    "    write_graph=True,\n",
    "    write_grads=True, \n",
    "    write_images = True, \n",
    "    update_freq = 'epoch', \n",
    "    profile_batch=0\n",
    ")\n",
    "\n",
    "# Training logger\n",
    "csv_log = os.path.join(logdir, 'training.csv')\n",
    "csv_logger = CSVLogger(csv_log, separator=',', append=True)\n",
    "\n",
    "# Only save the best model weights based on the val_loss\n",
    "checkpoint = ModelCheckpoint(os.path.join(logdir, 'snn_model-{epoch:02d}-{val_loss:.2f}.h5'),\n",
    "                             monitor='val_loss', verbose=1,\n",
    "                             save_best_only=True, save_weights_only=True, \n",
    "                             mode='auto')\n",
    "\n",
    "# Save the embedding mode weights based on the main model's val loss\n",
    "# This is needed to reecreate the emebedding model should we wish to visualise\n",
    "# the latent space at the saved epoch\n",
    "class SaveEmbeddingModelWeights(Callback):\n",
    "    def __init__(self, filepath, monitor='val_loss', verbose=1):\n",
    "        super(Callback, self).__init__()\n",
    "        self.monitor = monitor\n",
    "        self.verbose = verbose\n",
    "        self.best = np.Inf\n",
    "        self.filepath = filepath\n",
    "\n",
    "    def on_epoch_end(self, epoch, logs={}):\n",
    "        current = logs.get(self.monitor)\n",
    "        if current is None:\n",
    "            warnings.warn(\"SaveEmbeddingModelWeights requires %s available!\" % self.monitor, RuntimeWarning)\n",
    "\n",
    "        if current < self.best:\n",
    "            filepath = self.filepath.format(epoch=epoch + 1, **logs)\n",
    "            #if self.verbose == 1:\n",
    "                #print(\"Saving embedding model weights at %s\" % filepath)\n",
    "            embedding_model.save_weights(filepath, overwrite = True)\n",
    "            self.best = current\n",
    "\n",
    "# Save the embedding model weights if you save a new snn best model based on the model checkpoint above\n",
    "emb_weight_saver = SaveEmbeddingModelWeights(os.path.join(logdir, 'emb_model-{epoch:02d}.h5'))\n",
    "\n",
    "\n",
    "callbacks = [tensorboard, csv_logger, checkpoint, emb_weight_saver]\n",
    "\n",
    "\n",
    "# Save model configs to JSON\n",
    "model_json = siamese_net.to_json()\n",
    "with open(os.path.join(logdir, \"siamese_config.json\"), \"w\") as json_file:\n",
    "    json_file.write(model_json)\n",
    "    json_file.close()\n",
    "    \n",
    "model_json = embedding_model.to_json()\n",
    "with open(os.path.join(logdir, \"embedding_config.json\"), \"w\") as json_file:\n",
    "    json_file.write(model_json)\n",
    "    json_file.close()\n",
    "    \n",
    "\n",
    "hyperparams = {'batch_size' : batch_size,\n",
    "              'epochs' : epochs, \n",
    "               'steps_per_epoch' : steps_per_epoch, \n",
    "               'val_steps' : val_steps, \n",
    "               'alpha' : alpha, \n",
    "               'num_hard' : num_hard, \n",
    "               'optimiser' : optimiser,\n",
    "               'lr' : lr,\n",
    "               'emb_size' : emb_size\n",
    "              }\n",
    "\n",
    "\n",
    "with open(os.path.join(logdir, \"hyperparams.json\"), \"w\") as json_file:\n",
    "    json.dump(hyperparams, json_file)\n",
    "    \n",
    "# Set the model to TB\n",
    "tensorboard.set_model(siamese_net)\n",
    "\n",
    "\n",
    "def delete_older_model_files(filepath):\n",
    "    \n",
    "    model_dir = filepath.split(\"emb_model\")[0]\n",
    "    \n",
    "    # Get model files\n",
    "    model_files = os.listdir(model_dir)\n",
    "\n",
    "    # Get only the emb_model files\n",
    "    emb_model_files = [file for file in model_files if \"emb_model\" in file]\n",
    "    # Get the epoch nums of the emb_model_files\n",
    "    emb_model_files_epoch_nums = [int(file.split(\"-\")[1].split(\".h5\")[0]) for file in emb_model_files]\n",
    "\n",
    "    # Find all the snn model files\n",
    "    snn_model_files = [file for file in model_files if \"snn_model\" in file]\n",
    "\n",
    "    # Sort, get highest epoch num\n",
    "    emb_model_files_epoch_nums.sort()\n",
    "    highest_epoch_num = str(emb_model_files_epoch_nums[-1]).zfill(2)\n",
    "\n",
    "    # Filter the emb_model and snn_model file lists to remove the highest epoch number ones\n",
    "    emb_model_files_without_highest = [file for file in emb_model_files if highest_epoch_num not in file]\n",
    "    snn_model_files_without_highest = [file for file in snn_model_files if (\"-\" + highest_epoch_num + \"-\") not in file]\n",
    "\n",
    "    # Delete the non-highest model files from the subdir\n",
    "    if len(emb_model_files_without_highest) != 0:\n",
    "        print(\"Deleting previous best model file\")\n",
    "    for model_file_list in [emb_model_files_without_highest, snn_model_files_without_highest]:\n",
    "        for file in model_file_list:\n",
    "            os.remove(os.path.join(model_dir, file))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Show example batches\n",
    "\n",
    "Based on code found [here](https://zhangruochi.com/Create-a-Siamese-Network-with-Triplet-Loss-in-Keras/2020/08/11/)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Example triplet batch:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x144 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Example semi-hard triplet batch:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x144 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Display sample batches. This has to be performed after the embedding model is created\n",
    "# as create_batch_hard utilises the model to see which batches are actually hard.\n",
    "\n",
    "examples = create_batch(1)\n",
    "print(\"Example triplet batch:\")\n",
    "plot_triplets(examples)\n",
    "\n",
    "print(\"Example semi-hard triplet batch:\")\n",
    "ex_hard = create_hard_batch(1, 1, split=\"train\")\n",
    "plot_triplets(ex_hard)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Training\n",
    "\n",
    "Using `.fit(workers = 0)` fixes the error when using hard batches where TF can't predict on the embedding network whilst fitting the siamese network (see: https://github.com/keras-team/keras/issues/5511#issuecomment-427666222)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Starting training process!\n",
      "-------------------------------------\n",
      "Epoch 1/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 6.9986\n",
      "Epoch 00001: val_loss improved from inf to 5.79736, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-01-5.80.h5\n",
      "234/234 [==============================] - 174s 742ms/step - loss: 6.9935 - val_loss: 5.7974\n",
      "Epoch 2/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 4.8596\n",
      "Epoch 00002: val_loss improved from 5.79736 to 4.00626, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-02-4.01.h5\n",
      "234/234 [==============================] - 170s 728ms/step - loss: 4.8560 - val_loss: 4.0063\n",
      "Epoch 3/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 3.3463\n",
      "Epoch 00003: val_loss improved from 4.00626 to 2.74879, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-03-2.75.h5\n",
      "234/234 [==============================] - 170s 725ms/step - loss: 3.3438 - val_loss: 2.7488\n",
      "Epoch 4/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 2.2919\n",
      "Epoch 00004: val_loss improved from 2.74879 to 1.87885, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-04-1.88.h5\n",
      "234/234 [==============================] - 170s 726ms/step - loss: 2.2901 - val_loss: 1.8789\n",
      "Epoch 5/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 1.5656\n",
      "Epoch 00005: val_loss improved from 1.87885 to 1.28401, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-05-1.28.h5\n",
      "234/234 [==============================] - 172s 733ms/step - loss: 1.5644 - val_loss: 1.2840\n",
      "Epoch 6/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 1.0720\n",
      "Epoch 00006: val_loss improved from 1.28401 to 0.88143, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-06-0.88.h5\n",
      "234/234 [==============================] - 174s 742ms/step - loss: 1.0712 - val_loss: 0.8814\n",
      "Epoch 7/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.7392\n",
      "Epoch 00007: val_loss improved from 0.88143 to 0.61174, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-07-0.61.h5\n",
      "234/234 [==============================] - 170s 728ms/step - loss: 0.7387 - val_loss: 0.6117\n",
      "Epoch 8/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.5158\n",
      "Epoch 00008: val_loss improved from 0.61174 to 0.43090, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-08-0.43.h5\n",
      "234/234 [==============================] - 171s 729ms/step - loss: 0.5155 - val_loss: 0.4309\n",
      "Epoch 9/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.3670\n",
      "Epoch 00009: val_loss improved from 0.43090 to 0.30955, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-09-0.31.h5\n",
      "234/234 [==============================] - 171s 731ms/step - loss: 0.3667 - val_loss: 0.3095\n",
      "Epoch 10/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.2651\n",
      "Epoch 00010: val_loss improved from 0.30955 to 0.22544, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-10-0.23.h5\n",
      "234/234 [==============================] - 167s 715ms/step - loss: 0.2649 - val_loss: 0.2254\n",
      "Epoch 11/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.1966\n",
      "Epoch 00011: val_loss improved from 0.22544 to 0.17017, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-11-0.17.h5\n",
      "234/234 [==============================] - 166s 711ms/step - loss: 0.1965 - val_loss: 0.1702\n",
      "Epoch 12/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.1488\n",
      "Epoch 00012: val_loss improved from 0.17017 to 0.13102, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-12-0.13.h5\n",
      "234/234 [==============================] - 167s 713ms/step - loss: 0.1488 - val_loss: 0.1310\n",
      "Epoch 13/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.1153\n",
      "Epoch 00013: val_loss improved from 0.13102 to 0.10136, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-13-0.10.h5\n",
      "234/234 [==============================] - 169s 722ms/step - loss: 0.1153 - val_loss: 0.1014\n",
      "Epoch 14/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0924\n",
      "Epoch 00014: val_loss improved from 0.10136 to 0.08184, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-14-0.08.h5\n",
      "234/234 [==============================] - 168s 718ms/step - loss: 0.0923 - val_loss: 0.0818\n",
      "Epoch 15/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0752\n",
      "Epoch 00015: val_loss improved from 0.08184 to 0.06727, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-15-0.07.h5\n",
      "234/234 [==============================] - 167s 714ms/step - loss: 0.0752 - val_loss: 0.0673\n",
      "Epoch 16/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0640\n",
      "Epoch 00016: val_loss improved from 0.06727 to 0.05794, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-16-0.06.h5\n",
      "234/234 [==============================] - 167s 712ms/step - loss: 0.0640 - val_loss: 0.0579\n",
      "Epoch 17/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0548\n",
      "Epoch 00017: val_loss improved from 0.05794 to 0.05117, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-17-0.05.h5\n",
      "234/234 [==============================] - 169s 720ms/step - loss: 0.0548 - val_loss: 0.0512\n",
      "Epoch 18/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0485\n",
      "Epoch 00018: val_loss improved from 0.05117 to 0.04612, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-18-0.05.h5\n",
      "234/234 [==============================] - 167s 714ms/step - loss: 0.0485 - val_loss: 0.0461\n",
      "Epoch 19/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0439\n",
      "Epoch 00019: val_loss improved from 0.04612 to 0.04303, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-19-0.04.h5\n",
      "234/234 [==============================] - 168s 716ms/step - loss: 0.0438 - val_loss: 0.0430\n",
      "Epoch 20/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0412\n",
      "Epoch 00020: val_loss improved from 0.04303 to 0.04107, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-20-0.04.h5\n",
      "234/234 [==============================] - 169s 723ms/step - loss: 0.0412 - val_loss: 0.0411\n",
      "Epoch 21/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0390\n",
      "Epoch 00021: val_loss improved from 0.04107 to 0.03768, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-21-0.04.h5\n",
      "234/234 [==============================] - 168s 718ms/step - loss: 0.0390 - val_loss: 0.0377\n",
      "Epoch 22/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0367\n",
      "Epoch 00022: val_loss improved from 0.03768 to 0.03581, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-22-0.04.h5\n",
      "234/234 [==============================] - 167s 715ms/step - loss: 0.0367 - val_loss: 0.0358\n",
      "Epoch 23/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0357\n",
      "Epoch 00023: val_loss did not improve from 0.03581\n",
      "234/234 [==============================] - 168s 720ms/step - loss: 0.0357 - val_loss: 0.0358\n",
      "Epoch 24/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0349\n",
      "Epoch 00024: val_loss improved from 0.03581 to 0.03299, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-24-0.03.h5\n",
      "234/234 [==============================] - 168s 716ms/step - loss: 0.0349 - val_loss: 0.0330\n",
      "Epoch 25/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0342\n",
      "Epoch 00025: val_loss did not improve from 0.03299\n",
      "234/234 [==============================] - 164s 700ms/step - loss: 0.0342 - val_loss: 0.0337\n",
      "Epoch 26/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0329\n",
      "Epoch 00026: val_loss did not improve from 0.03299\n",
      "234/234 [==============================] - 166s 711ms/step - loss: 0.0330 - val_loss: 0.0353\n",
      "Epoch 27/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0330\n",
      "Epoch 00027: val_loss improved from 0.03299 to 0.03299, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-27-0.03.h5\n",
      "234/234 [==============================] - 169s 723ms/step - loss: 0.0330 - val_loss: 0.0330\n",
      "Epoch 28/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0328\n",
      "Epoch 00028: val_loss did not improve from 0.03299\n",
      "234/234 [==============================] - 167s 716ms/step - loss: 0.0328 - val_loss: 0.0335\n",
      "Epoch 29/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0328\n",
      "Epoch 00029: val_loss improved from 0.03299 to 0.03119, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-29-0.03.h5\n",
      "234/234 [==============================] - 168s 719ms/step - loss: 0.0328 - val_loss: 0.0312\n",
      "Epoch 30/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0318\n",
      "Epoch 00030: val_loss improved from 0.03119 to 0.03046, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-30-0.03.h5\n",
      "234/234 [==============================] - 167s 714ms/step - loss: 0.0319 - val_loss: 0.0305\n",
      "Epoch 31/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0317\n",
      "Epoch 00031: val_loss did not improve from 0.03046\n",
      "234/234 [==============================] - 168s 719ms/step - loss: 0.0317 - val_loss: 0.0320\n",
      "Epoch 32/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0312\n",
      "Epoch 00032: val_loss did not improve from 0.03046\n",
      "234/234 [==============================] - 165s 703ms/step - loss: 0.0312 - val_loss: 0.0316\n",
      "Epoch 33/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0312\n",
      "Epoch 00033: val_loss improved from 0.03046 to 0.03021, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-33-0.03.h5\n",
      "234/234 [==============================] - 170s 726ms/step - loss: 0.0312 - val_loss: 0.0302\n",
      "Epoch 34/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0317\n",
      "Epoch 00034: val_loss did not improve from 0.03021\n",
      "234/234 [==============================] - 170s 727ms/step - loss: 0.0317 - val_loss: 0.0313\n",
      "Epoch 35/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0315\n",
      "Epoch 00035: val_loss did not improve from 0.03021\n",
      "234/234 [==============================] - 167s 715ms/step - loss: 0.0315 - val_loss: 0.0307\n",
      "Epoch 36/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0313\n",
      "Epoch 00036: val_loss improved from 0.03021 to 0.02951, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-36-0.03.h5\n",
      "234/234 [==============================] - 169s 722ms/step - loss: 0.0313 - val_loss: 0.0295\n",
      "Epoch 37/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0312\n",
      "Epoch 00037: val_loss did not improve from 0.02951\n",
      "234/234 [==============================] - 166s 710ms/step - loss: 0.0312 - val_loss: 0.0305\n",
      "Epoch 38/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0307\n",
      "Epoch 00038: val_loss did not improve from 0.02951\n",
      "234/234 [==============================] - 168s 719ms/step - loss: 0.0307 - val_loss: 0.0304\n",
      "Epoch 39/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0308\n",
      "Epoch 00039: val_loss did not improve from 0.02951\n",
      "234/234 [==============================] - 165s 707ms/step - loss: 0.0308 - val_loss: 0.0318\n",
      "Epoch 40/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0312\n",
      "Epoch 00040: val_loss did not improve from 0.02951\n",
      "234/234 [==============================] - 168s 719ms/step - loss: 0.0312 - val_loss: 0.0305\n",
      "Epoch 41/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0309\n",
      "Epoch 00041: val_loss did not improve from 0.02951\n",
      "234/234 [==============================] - 166s 711ms/step - loss: 0.0309 - val_loss: 0.0303\n",
      "Epoch 42/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0305\n",
      "Epoch 00042: val_loss did not improve from 0.02951\n",
      "234/234 [==============================] - 164s 700ms/step - loss: 0.0305 - val_loss: 0.0296\n",
      "Epoch 43/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0307\n",
      "Epoch 00043: val_loss did not improve from 0.02951\n",
      "234/234 [==============================] - 166s 710ms/step - loss: 0.0307 - val_loss: 0.0301\n",
      "Epoch 44/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0308\n",
      "Epoch 00044: val_loss did not improve from 0.02951\n",
      "234/234 [==============================] - 169s 721ms/step - loss: 0.0308 - val_loss: 0.0299\n",
      "Epoch 45/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0305\n",
      "Epoch 00045: val_loss did not improve from 0.02951\n",
      "234/234 [==============================] - 169s 722ms/step - loss: 0.0305 - val_loss: 0.0307\n",
      "Epoch 46/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0309\n",
      "Epoch 00046: val_loss did not improve from 0.02951\n",
      "234/234 [==============================] - 168s 719ms/step - loss: 0.0309 - val_loss: 0.0298\n",
      "Epoch 47/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0305\n",
      "Epoch 00047: val_loss did not improve from 0.02951\n",
      "234/234 [==============================] - 168s 719ms/step - loss: 0.0306 - val_loss: 0.0299\n",
      "Epoch 48/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0302\n",
      "Epoch 00048: val_loss did not improve from 0.02951\n",
      "234/234 [==============================] - 167s 713ms/step - loss: 0.0302 - val_loss: 0.0302\n",
      "Epoch 49/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0307\n",
      "Epoch 00049: val_loss improved from 0.02951 to 0.02931, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-49-0.03.h5\n",
      "234/234 [==============================] - 169s 720ms/step - loss: 0.0307 - val_loss: 0.0293\n",
      "Epoch 50/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0306\n",
      "Epoch 00050: val_loss did not improve from 0.02931\n",
      "234/234 [==============================] - 167s 714ms/step - loss: 0.0306 - val_loss: 0.0299\n",
      "Epoch 51/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0302\n",
      "Epoch 00051: val_loss improved from 0.02931 to 0.02882, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-51-0.03.h5\n",
      "234/234 [==============================] - 167s 712ms/step - loss: 0.0302 - val_loss: 0.0288\n",
      "Epoch 52/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0308\n",
      "Epoch 00052: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 166s 710ms/step - loss: 0.0308 - val_loss: 0.0296\n",
      "Epoch 53/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0304\n",
      "Epoch 00053: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 167s 714ms/step - loss: 0.0304 - val_loss: 0.0303\n",
      "Epoch 54/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0304\n",
      "Epoch 00054: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 167s 712ms/step - loss: 0.0304 - val_loss: 0.0299\n",
      "Epoch 55/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0300\n",
      "Epoch 00055: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 166s 707ms/step - loss: 0.0300 - val_loss: 0.0294\n",
      "Epoch 56/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0304\n",
      "Epoch 00056: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 167s 715ms/step - loss: 0.0304 - val_loss: 0.0300\n",
      "Epoch 57/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0306\n",
      "Epoch 00057: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 167s 713ms/step - loss: 0.0306 - val_loss: 0.0295\n",
      "Epoch 58/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "233/234 [============================>.] - ETA: 0s - loss: 0.0309\n",
      "Epoch 00058: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 165s 706ms/step - loss: 0.0309 - val_loss: 0.0296\n",
      "Epoch 59/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0303\n",
      "Epoch 00059: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 167s 714ms/step - loss: 0.0303 - val_loss: 0.0300\n",
      "Epoch 60/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0302\n",
      "Epoch 00060: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 168s 716ms/step - loss: 0.0302 - val_loss: 0.0294\n",
      "Epoch 61/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0298\n",
      "Epoch 00061: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 166s 709ms/step - loss: 0.0298 - val_loss: 0.0303\n",
      "Epoch 62/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0301\n",
      "Epoch 00062: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 168s 717ms/step - loss: 0.0302 - val_loss: 0.0301\n",
      "Epoch 63/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0304\n",
      "Epoch 00063: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 167s 715ms/step - loss: 0.0304 - val_loss: 0.0303\n",
      "Epoch 64/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0302\n",
      "Epoch 00064: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 167s 712ms/step - loss: 0.0301 - val_loss: 0.0290\n",
      "Epoch 65/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0301\n",
      "Epoch 00065: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 164s 701ms/step - loss: 0.0301 - val_loss: 0.0305\n",
      "Epoch 66/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0313\n",
      "Epoch 00066: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 167s 715ms/step - loss: 0.0313 - val_loss: 0.0300\n",
      "Epoch 67/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0305\n",
      "Epoch 00067: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 169s 720ms/step - loss: 0.0305 - val_loss: 0.0304\n",
      "Epoch 68/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0300\n",
      "Epoch 00068: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 166s 710ms/step - loss: 0.0300 - val_loss: 0.0289\n",
      "Epoch 69/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0298\n",
      "Epoch 00069: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 166s 708ms/step - loss: 0.0298 - val_loss: 0.0300\n",
      "Epoch 70/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0299\n",
      "Epoch 00070: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 168s 718ms/step - loss: 0.0299 - val_loss: 0.0304\n",
      "Epoch 71/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0306\n",
      "Epoch 00071: val_loss did not improve from 0.02882\n",
      "234/234 [==============================] - 166s 708ms/step - loss: 0.0307 - val_loss: 0.0319\n",
      "Epoch 72/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0306\n",
      "Epoch 00072: val_loss improved from 0.02882 to 0.02837, saving model to PATH/TO/LOGDIR/snn-example-run/snn_model-72-0.03.h5\n",
      "234/234 [==============================] - 169s 721ms/step - loss: 0.0306 - val_loss: 0.0284\n",
      "Epoch 73/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0301\n",
      "Epoch 00073: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 167s 713ms/step - loss: 0.0301 - val_loss: 0.0303\n",
      "Epoch 74/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0303\n",
      "Epoch 00074: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 167s 713ms/step - loss: 0.0303 - val_loss: 0.0305\n",
      "Epoch 75/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0302\n",
      "Epoch 00075: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 167s 712ms/step - loss: 0.0302 - val_loss: 0.0307\n",
      "Epoch 76/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0300\n",
      "Epoch 00076: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 166s 709ms/step - loss: 0.0299 - val_loss: 0.0307\n",
      "Epoch 77/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0300\n",
      "Epoch 00077: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 166s 710ms/step - loss: 0.0301 - val_loss: 0.0288\n",
      "Epoch 78/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0301\n",
      "Epoch 00078: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 169s 721ms/step - loss: 0.0301 - val_loss: 0.0298\n",
      "Epoch 79/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0295\n",
      "Epoch 00079: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 166s 709ms/step - loss: 0.0295 - val_loss: 0.0298\n",
      "Epoch 80/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0300\n",
      "Epoch 00080: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 167s 712ms/step - loss: 0.0300 - val_loss: 0.0293\n",
      "Epoch 81/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0298\n",
      "Epoch 00081: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 168s 718ms/step - loss: 0.0298 - val_loss: 0.0309\n",
      "Epoch 82/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0301\n",
      "Epoch 00082: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 166s 709ms/step - loss: 0.0301 - val_loss: 0.0295\n",
      "Epoch 83/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0299\n",
      "Epoch 00083: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 166s 710ms/step - loss: 0.0299 - val_loss: 0.0303\n",
      "Epoch 84/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0301\n",
      "Epoch 00084: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 166s 710ms/step - loss: 0.0301 - val_loss: 0.0299\n",
      "Epoch 85/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0301\n",
      "Epoch 00085: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 168s 717ms/step - loss: 0.0301 - val_loss: 0.0301\n",
      "Epoch 86/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0297\n",
      "Epoch 00086: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 167s 712ms/step - loss: 0.0297 - val_loss: 0.0296\n",
      "Epoch 87/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0303\n",
      "Epoch 00087: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 166s 708ms/step - loss: 0.0303 - val_loss: 0.0302\n",
      "Epoch 88/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0295\n",
      "Epoch 00088: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 165s 707ms/step - loss: 0.0295 - val_loss: 0.0297\n",
      "Epoch 89/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0296\n",
      "Epoch 00089: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 168s 716ms/step - loss: 0.0296 - val_loss: 0.0287\n",
      "Epoch 90/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0301\n",
      "Epoch 00090: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 168s 718ms/step - loss: 0.0301 - val_loss: 0.0293\n",
      "Epoch 91/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0299\n",
      "Epoch 00091: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 167s 715ms/step - loss: 0.0299 - val_loss: 0.0297\n",
      "Epoch 92/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0295\n",
      "Epoch 00092: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 168s 720ms/step - loss: 0.0295 - val_loss: 0.0289\n",
      "Epoch 93/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0303\n",
      "Epoch 00093: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 172s 736ms/step - loss: 0.0302 - val_loss: 0.0288\n",
      "Epoch 94/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0302\n",
      "Epoch 00094: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 168s 719ms/step - loss: 0.0302 - val_loss: 0.0296\n",
      "Epoch 95/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0301\n",
      "Epoch 00095: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 170s 727ms/step - loss: 0.0301 - val_loss: 0.0296\n",
      "Epoch 96/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0299\n",
      "Epoch 00096: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 170s 725ms/step - loss: 0.0298 - val_loss: 0.0306\n",
      "Epoch 97/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0302\n",
      "Epoch 00097: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 168s 719ms/step - loss: 0.0302 - val_loss: 0.0288\n",
      "Epoch 98/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0301\n",
      "Epoch 00098: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 167s 714ms/step - loss: 0.0301 - val_loss: 0.0290\n",
      "Epoch 99/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0299\n",
      "Epoch 00099: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 167s 714ms/step - loss: 0.0299 - val_loss: 0.0289\n",
      "Epoch 100/100\n",
      "233/234 [============================>.] - ETA: 0s - loss: 0.0305\n",
      "Epoch 00100: val_loss did not improve from 0.02837\n",
      "234/234 [==============================] - 167s 714ms/step - loss: 0.0305 - val_loss: 0.0297\n",
      "-------------------------------------\n",
      "Training complete.\n"
     ]
    }
   ],
   "source": [
    "def get_num_gpus():\n",
    "    local_device_protos = device_lib.list_local_devices()\n",
    "    return len([x.name for x in local_device_protos if x.device_type == 'GPU'])\n",
    "\n",
    "## Training:\n",
    "#print(\"Logging out to Tensorboard at:\", logdir)\n",
    "print(\"Starting training process!\")\n",
    "print(\"-------------------------------------\")\n",
    "\n",
    "# Make the model work over the two GPUs we have\n",
    "num_gpus = get_num_gpus()\n",
    "parallel_snn = multi_gpu_model(siamese_net, gpus = num_gpus)\n",
    "batch_per_gpu = int(batch_size / num_gpus)\n",
    "\n",
    "parallel_snn.compile(loss=triplet_loss, optimizer= optimiser_obj)\n",
    "\n",
    "siamese_history = parallel_snn.fit(\n",
    "    data_generator(batch_per_gpu, num_hard),\n",
    "    steps_per_epoch=steps_per_epoch,\n",
    "    epochs=epochs,\n",
    "    verbose=1,\n",
    "    callbacks=callbacks, \n",
    "    workers = 0, \n",
    "    validation_data = data_generator(batch_per_gpu, num_hard, split=\"test\"), \n",
    "    validation_steps = val_steps)\n",
    "\n",
    "print(\"-------------------------------------\")\n",
    "print(\"Training complete.\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluate the trained network\n",
    "\n",
    "Load the best performing models. We need to load the weights and configs seperately rather than using model.load() as our custom loss function relies on the embedding length. As such, it is easier to load the weights and config seperately and build a model based on them."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "def json_to_dict(json_src):\n",
    "    with open(json_src, 'r') as j:\n",
    "        return json.loads(j.read())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x144 with 11 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x144 with 11 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x144 with 11 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Load in best trained SNN and emb model\n",
    "\n",
    "# The best performing model weights has the higher epoch number due to only saving the best weights\n",
    "highest_epoch = 0\n",
    "dir_list = os.listdir(logdir)\n",
    "\n",
    "for file in dir_list:\n",
    "    if file.endswith(\".h5\"):\n",
    "        epoch_num = int(file.split(\"-\")[1].split(\".h5\")[0])\n",
    "        if epoch_num > highest_epoch:\n",
    "            highest_epoch = epoch_num\n",
    "\n",
    "# Find the embedding and SNN weights src for the highest_epoch (best) model\n",
    "for file in dir_list:\n",
    "    # Zfill ensure a leading 0 on number < 10\n",
    "    if (\"-\" + str(highest_epoch).zfill(2)) in file:\n",
    "        if file.startswith(\"emb\"):\n",
    "            embedding_weights_src = os.path.join(logdir, file)\n",
    "        elif file.startswith(\"snn\"):\n",
    "            snn_weights_src = os.path.join(logdir, file)\n",
    "\n",
    "hyperparams = os.path.join(logdir, \"hyperparams.json\")\n",
    "snn_config = os.path.join(logdir, \"siamese_config.json\")\n",
    "emb_config = os.path.join(logdir, \"embedding_config.json\")\n",
    "\n",
    "snn_config = json_to_dict(snn_config)\n",
    "emb_config = json_to_dict(emb_config)\n",
    "\n",
    "# json.dumps to make the dict a string, as required by model_from_json\n",
    "loaded_snn_model = model_from_json(json.dumps(snn_config))\n",
    "loaded_snn_model.load_weights(snn_weights_src)\n",
    "\n",
    "loaded_emb_model = model_from_json(json.dumps(emb_config))\n",
    "loaded_emb_model.load_weights(embedding_weights_src)\n",
    "\n",
    "\n",
    "# Store visualisations of the embeddings using PCA for display next to \"after training\" for comparisons\n",
    "embeddings_after_train = loaded_emb_model.predict(x_test[:num_vis, :])\n",
    "pca = PCA(n_components=2)\n",
    "decomposed_embeddings_after = pca.fit_transform(embeddings_after_train)\n",
    "evaluate(loaded_emb_model, highest_epoch)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Comparisons of the embeddings in the latent space\n",
    "\n",
    "Based on [this notebook](https://github.com/AdrianUng/keras-triplet-loss-mnist/blob/master/Triplet_loss_KERAS_semi_hard_from_TF.ipynb)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "step = 1 # Step = 1, take every element\n",
    "\n",
    "dict_embeddings = {}\n",
    "dict_gray = {}\n",
    "test_class_labels = np.unique(np.array(y_test))\n",
    "\n",
    "decomposed_embeddings_after = pca.fit_transform(embeddings_after_train)\n",
    "\n",
    "fig = plt.figure(figsize=(16, 8))\n",
    "for label in test_class_labels:\n",
    "    \n",
    "    y_test_labels = y_test[:num_vis]\n",
    "    \n",
    "    decomposed_embeddings_class_before = decomposed_embeddings_before[y_test_labels == label]\n",
    "    decomposed_embeddings_class_after = decomposed_embeddings_after[y_test_labels == label]\n",
    "    \n",
    "    plt.subplot(1,2,1)\n",
    "    plt.scatter(decomposed_embeddings_class_before[::step, 1], decomposed_embeddings_class_before[::step, 0], label=str(label))\n",
    "    plt.title('Embedding Locations Before Training')\n",
    "    plt.legend()\n",
    "\n",
    "    plt.subplot(1,2,2)\n",
    "    plt.scatter(decomposed_embeddings_class_after[::step, 1], decomposed_embeddings_class_after[::step, 0], label=str(label))\n",
    "    plt.title('Embedding Locations After %d Training Epochs' % epochs)\n",
    "    plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Determine n_way_accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "97.43589743589743"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prototypes = generate_prototypes(x_test, y_test, loaded_emb_model)\n",
    "n_way_accuracy_prototypes(val_steps, num_classes, loaded_emb_model)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Visualise support set inference\n",
    "\n",
    "Based on code found [here](https://github.com/asagar60/One-Shot-Learning/blob/master/Omniglot_data/One_shot_implementation.ipynb)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Image most similar to Test Image in Support Set')"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x864 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "n_samples = 10\n",
    "sample_imgs, min_index = visualise_n_way_prototypes(n_samples, loaded_emb_model)\n",
    "\n",
    "img_matrix = []\n",
    "for index in range(1, len(sample_imgs)):\n",
    "    img_matrix.append(np.reshape(sample_imgs[index], (x_train_w, x_train_h)))\n",
    "\n",
    "img_matrix = np.asarray(img_matrix)\n",
    "img_matrix = np.vstack(img_matrix)\n",
    "\n",
    "f, ax = plt.subplots(1, 3, figsize = (10, 12))\n",
    "f.tight_layout()\n",
    "ax[0].imshow(np.reshape(sample_imgs[0], (x_train_w, x_train_h)),vmin=0, vmax=1,cmap='Greys')\n",
    "ax[0].set_title(\"Test Image\")\n",
    "ax[1].imshow(img_matrix ,vmin=0, vmax=1,cmap='Greys')\n",
    "ax[1].set_title(\"Support Set (Img of same class shown first)\")\n",
    "ax[2].imshow(np.reshape(sample_imgs[min_index], (x_train_w, x_train_h)),vmin=0, vmax=1,cmap='Greys')\n",
    "ax[2].set_title(\"Image most similar to Test Image in Support Set\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.9.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}