Siamese_Keras_InceptionModel.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Siamese_Keras_InceptionModel",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "UCGHZG-MhL8k",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "import os\n",
        "import h5py\n",
        "import tensorflow as tf\n",
        "import keras\n",
        "import PIL\n",
        "import scipy\n",
        "from keras.layers import *\n",
        "from keras.regularizers import *\n",
        "from keras.optimizers import *\n",
        "from keras.models import *\n",
        "from keras.initializers import *\n",
        "from keras.backend import *\n",
        "\n",
        "%matplotlib inline"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CCw7T93SicfJ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "os.chdir('drive/My Drive/Dataset')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oaLp-RUKiwPc",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def Load_Data():\n",
        "  \n",
        "  train_dataset = h5py.File('train_happy.h5','r')\n",
        "  test_dataset = h5py.File('test_happy.h5','r')\n",
        "  \n",
        "  train_set_X_orig = np.array(train_dataset['train_set_x'][:])\n",
        "  train_set_Y_orig = np.array(train_dataset['train_set_y'][:])\n",
        "  test_set_X_orig = np.array(test_dataset['test_set_x'][:])\n",
        "  test_set_Y_orig = np.array(test_dataset['test_set_y'][:])\n",
        "  \n",
        "  classes = np.array(test_dataset['list_classes'][:])\n",
        "  \n",
        "  return train_set_X_orig,train_set_Y_orig,test_set_X_orig,test_set_Y_orig,classes"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fLoI2o_Sk7vN",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 305
        },
        "outputId": "66d90261-e299-4965-9859-401e64ddb21b"
      },
      "source": [
        "train_set_X_orig,train_set_Y_orig,test_set_X_orig,test_set_Y_orig,classes = Load_Data()\n",
        "print(train_set_X_orig.shape,train_set_Y_orig.shape,test_set_X_orig.shape,test_set_Y_orig.shape)\n",
        "i = 5\n",
        "plt.imshow(train_set_X_orig[i])\n",
        "print('It is Image_'+str(i)+' and the person is','Happy.' if classes[test_set_Y_orig[i]] else 'not Happy.')"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(600, 64, 64, 3) (600,) (150, 64, 64, 3) (150,)\n",
            "It is Image_5 and the person is Happy.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD8CAYAAABXXhlaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n38NjCuEXvofHFOL/B9wjTh4Mp7aiAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "z7baZt17lA_7",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def Reshape_And_Normalize(train_set_X_orig,train_set_Y_orig,test_set_X_orig,test_set_Y_orig):\n",
        " \n",
        "  train_set_X = train_set_X_orig/255\n",
        "  train_set_Y = train_set_Y_orig.reshape((train_set_Y_orig.shape[0],1))\n",
        "  test_set_X = test_set_X_orig/255\n",
        "  test_set_Y = test_set_Y_orig.reshape((test_set_Y_orig.shape[0],1))\n",
        "  \n",
        "  return train_set_X,train_set_Y,test_set_X,test_set_Y"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EHXb0vDaSNE6",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def Read_Outputs():\n",
        "  \n",
        "  ptr = open('outputs_siamese.txt','r')\n",
        "  data = ptr.readlines()\n",
        "\n",
        "  train_Y = [float(c.strip()) for c in data]\n",
        "  train_Y = np.array(train_Y)\n",
        "  train_Y = train_Y.reshape((train_Y.shape[0],1))\n",
        "  \n",
        "  return train_Y"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kMqOrvrGZElJ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def Create_Database():\n",
        "  \n",
        "  database = {}\n",
        "  database[\"younes\"] = \"images/younes.jpg\"\n",
        "  database[\"tian\"] = \"images/tian.jpg\"\n",
        "  database[\"andrew\"] = \"images/andrew.jpg\"\n",
        "  database[\"kian\"] = \"images/kian.jpg\"\n",
        "  database[\"dan\"] = \"images/dan.jpg\"\n",
        "  database[\"sebastiano\"] = \"images/sebastiano.jpg\"\n",
        "  database[\"bertrand\"] = \"images/bertrand.jpg\"\n",
        "  database[\"kevin\"] = \"images/kevin.jpg\"\n",
        "  database[\"felix\"] = \"images/felix.jpg\"\n",
        "  database[\"benoit\"] = \"images/benoit.jpg\"\n",
        "  database[\"arnaud\"] = \"images/arnaud.jpg\"\n",
        "  \n",
        "  return database"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JgA-7mhjGG2o",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def Data_Preprocessing():\n",
        "\n",
        "  train_set_X_orig,train_set_Y_orig,test_set_X_orig,test_set_Y_orig,classes = Load_Data()\n",
        "  train_set_X,train_set_Y,test_set_X,test_set_Y = Reshape_And_Normalize(train_set_X_orig,train_set_Y_orig,test_set_X_orig,test_set_Y_orig)\n",
        "  \n",
        "  data = train_set_X_orig[0:21,:,:,:]\n",
        "  temp = np.zeros((1,64,64,3))\n",
        "  temp[0,:,:,:] = data[1,:,:,:]\n",
        "  data[1,:,:,:] = data[2,:,:,:]\n",
        "  data[2,:,:,:] = temp[0,:,:,:]\n",
        "  temp[0,:,:,:] = data[15,:,:,:]\n",
        "  data[15,:,:,:] = data[19,:,:,:]\n",
        "  data[19,:,:,:] = temp[0,:,:,:]\n",
        "  temp[0,:,:,:] = data[16,:,:,:]\n",
        "  data[16,:,:,:] = data[17,:,:,:]\n",
        "  data[17,:,:,:] = temp[0,:,:,:]\n",
        "  \n",
        "  train_X1 = data[0:20,:,:,:]\n",
        "  train_X2 = data[1:21,:,:,:]\n",
        "  \n",
        "  train_Y = Read_Outputs()\n",
        "  database = Create_Database()\n",
        "  \n",
        "  return train_X1,train_X2,train_Y,database"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Zi5mnumnnnjC",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def Inception_Block(X,channels,layers,XorY):\n",
        "  \n",
        "  X_3x3 = Conv2D(channels[0],kernel_size=(1,1),strides=(1,1),padding='valid',name=\"Inception_Block\"+str(layers)+\"_Conv2D_1x1A\"+XorY)(X)\n",
        "  X_3x3 = BatchNormalization(axis=3,name='Inception_Block'+str(layers)+'_BatchNorm_1'+XorY)(X_3x3)\n",
        "  X_3x3 = Activation('relu')(X_3x3)\n",
        "  X_3x3 = ZeroPadding2D((1,1))(X_3x3)\n",
        "  X_3x3 = Conv2D(channels[1],kernel_size=(3,3),strides=(1,1),padding='valid',name='Inception_Block'+str(layers)+'_Conv2D_3x3'+XorY)(X_3x3)\n",
        "  X_3x3 = BatchNormalization(axis=3,name='Inception_Block'+str(layers)+'_BatchNorm_2'+XorY)(X_3x3)\n",
        "  X_3x3 = Activation('relu')(X_3x3)\n",
        "  \n",
        "  \n",
        "  X_5x5 = Conv2D(channels[2],kernel_size=(1,1),strides=(1,1),padding='valid',name=\"Inception_Block\"+str(layers)+\"_Conv2D_1x1B\"+XorY)(X)\n",
        "  X_5x5 = BatchNormalization(axis=3,name='Inception_Block'+str(layers)+'_BatchNorm_3'+XorY)(X_5x5)\n",
        "  X_5x5 = Activation('relu')(X_5x5)\n",
        "  X_5x5 = ZeroPadding2D((2,2))(X_5x5)\n",
        "  X_5x5 = Conv2D(channels[3],kernel_size=(5,5),strides=(1,1),padding='valid',name='Inception_Block'+str(layers)+'_Conv2D_5x5'+XorY)(X_5x5)\n",
        "  X_5x5 = BatchNormalization(axis=3,name='Inception_Block'+str(layers)+'_BatchNorm_4'+XorY)(X_5x5)\n",
        "  X_5x5 = Activation('relu')(X_5x5)\n",
        "  \n",
        "\n",
        "  X_Pool = MaxPooling2D(pool_size=(2,2),strides=(1,1),padding='same',name='Inception_Block'+str(layers)+'_MaxPool'+XorY)(X)\n",
        "  X_Pool = Conv2D(channels[4],kernel_size=(1,1),strides=(1,1),padding='valid',name='Inception_Block'+str(layers)+'_Conv2D_1x1C'+XorY)(X_Pool)\n",
        "  X_Pool = BatchNormalization(axis=3,name=\"Inception_Block\"+str(layers)+\"_BatchNorm_6\"+XorY)(X_Pool)\n",
        "  X_Pool = Activation('relu')(X_Pool)\n",
        "  \n",
        "  \n",
        "  X_1x1 = Conv2D(channels[5],kernel_size=(1,1),strides=(1,1),padding='valid',name='Inception_Block'+str(layers)+'_Conv2D_1x1D'+XorY)(X)\n",
        "  X_1x1 = BatchNormalization(axis=3,name='Inception_Block'+str(layers)+'_BatchNorm_5'+XorY)(X_1x1)\n",
        "  X_1x1 = Activation('relu')(X_1x1)\n",
        "  \n",
        "  inception = Concatenate(axis=3)([X_3x3,X_5x5,X_Pool,X_1x1])\n",
        "  \n",
        "  return inception"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "O7QjgJ5h7IYa",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def Propagation(input_shape,output_neurons,channel_list):\n",
        "  \n",
        "  X_input = Input(input_shape)\n",
        "  X = X_input\n",
        "  \n",
        "  X = ZeroPadding2D((3,3))(X)\n",
        "  X = Conv2D(64,kernel_size=(3,3),strides=(1,1),padding='valid',kernel_initializer='glorot_uniform',name='Normal_Block_Conv2D_X')(X)\n",
        "  X = BatchNormalization(axis=3,name='Normal_Block_BatchNorm_X')(X)\n",
        "  X = Activation('relu')(X)\n",
        "  X = MaxPooling2D(pool_size=(2,2),strides=(2,2),padding='valid',name='Normal_Block_MaxPool1_X')(X)\n",
        "  \n",
        "  '''X = Inception_Block(X,channel_list[0],1,'X')\n",
        "  X = Inception_Block(X,channel_list[1],2,'X')\n",
        "  X = Inception_Block(X,channel_list[2],3,'X')\n",
        "  \n",
        "  X = MaxPooling2D(pool_size=(2,2),strides=(2,2),padding='valid',name=\"Normal_Block_MaxPool2_X\")(X)\n",
        "  X = Inception_Block(X,channel_list[3],4,'X')\n",
        "  X = Inception_Block(X,channel_list[4],5,'X')\n",
        "  \n",
        "  X = MaxPooling2D(pool_size=(2,2),strides=(2,2),padding='valid',name=\"Normal_Block_MaxPool3_X\")(X)\n",
        "  X = Inception_Block(X,channel_list[5],6,'X')\n",
        "  X = Inception_Block(X,channel_list[6],7,'X')'''\n",
        "  \n",
        "  X = AveragePooling2D(pool_size=(2,2),strides=(2,2),padding='valid',name='Normal_Block_AveragePool_X')(X)\n",
        "  X = Flatten()(X)\n",
        "  \n",
        "  if(output_neurons > 1):\n",
        "    activation = 'softmax'\n",
        "  else:\n",
        "    activation = 'sigmoid'\n",
        "  X = Dense(output_neurons,activation=activation,kernel_initializer='glorot_uniform',bias_initializer='zeros',name='Normal_Layer_FullyConnected_X',use_bias=True)(X)\n",
        "  \n",
        "  \n",
        "  Y_input = Input(input_shape)\n",
        "  Y = Y_input\n",
        "  \n",
        "  Y = ZeroPadding2D((3,3))(Y)\n",
        "  Y = Conv2D(64,kernel_size=(3,3),strides=(1,1),padding='valid',kernel_initializer='glorot_uniform',name='Normal_Block_Conv2D_Y')(Y)\n",
        "  Y = BatchNormalization(axis=3,name='Normal_Block_BatchNorm_Y')(Y)\n",
        "  Y = Activation('relu')(Y)\n",
        "  Y = MaxPooling2D(pool_size=(2,2),strides=(2,2),padding='valid',name='Normal_Block_MaxPool1_Y')(Y)\n",
        "  \n",
        "  '''Y = Inception_Block(Y,channel_list[0],1,'Y')\n",
        "  Y = Inception_Block(Y,channel_list[1],2,'Y')\n",
        "  Y = Inception_Block(Y,channel_list[2],3,'Y')\n",
        "  \n",
        "  Y = MaxPooling2D(pool_size=(2,2),strides=(2,2),padding='valid',name=\"Normal_Block_MaxPool2_Y\")(Y)\n",
        "  Y = Inception_Block(Y,channel_list[3],4,'Y')\n",
        "  Y = Inception_Block(Y,channel_list[4],5,'Y')\n",
        "  \n",
        "  Y = MaxPooling2D(pool_size=(2,2),strides=(2,2),padding='valid',name=\"Normal_Block_MaxPool3_Y\")(Y)\n",
        "  Y = Inception_Block(Y,channel_list[5],6,'Y')\n",
        "  Y = Inception_Block(Y,channel_list[6],7,'Y')'''\n",
        "  \n",
        "  Y = AveragePooling2D(pool_size=(2,2),strides=(2,2),padding='valid',name='Normal_Block_AveragePool_Y')(Y)\n",
        "  Y = Flatten()(Y)\n",
        "  \n",
        "  if(output_neurons > 1):\n",
        "    activation = 'softmax'\n",
        "  else:\n",
        "    activation = 'sigmoid'\n",
        "\n",
        "  Y = Dense(output_neurons,activation=activation,kernel_initializer='glorot_uniform',bias_initializer='zeros',name='Normal_Layer_FullyConnected_Y',use_bias=True)(Y)\n",
        "  \n",
        "  A = Subtract()([X,Y])\n",
        "  A = Dense(50,activation='softmax',kernel_initializer='glorot_uniform', bias_initializer='zeros')(A)\n",
        "  A = Dense(1,activation='sigmoid',name=\"Siamese_Last_Layer\",kernel_initializer='glorot_uniform', bias_initializer='zeros')(A)\n",
        "  model = Model(inputs=[X_input,Y_input],outputs=A,name='Siamese_Keras_InceptionModel')\n",
        "  \n",
        "  return model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UpPWeDwOiWpj",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def Read_Image(image_path):\n",
        "  \n",
        "  import skimage\n",
        "  \n",
        "  image = plt.imread(image_path)\n",
        "  image = skimage.transform.resize(image,(64,64))\n",
        "  image = np.expand_dims(image,axis=0)\n",
        "  \n",
        "  return image"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7LDg4tNNXaPq",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def Reshape_And_Normalize_Happy(train_set_X_orig,train_set_Y_orig,test_set_X_orig,test_set_Y_orig):\n",
        "  \n",
        "  m_train = train_set_X_orig.shape[0]\n",
        "  m_test = test_set_X_orig.shape[0]\n",
        "  train_set_X = train_set_X_orig.reshape(m_train,-1)\n",
        "  test_set_X = test_set_X_orig.reshape(m_test,-1)\n",
        "\n",
        "  train_set_Y = train_set_Y_orig.reshape((m_train,1))\n",
        "  test_set_Y = test_set_Y_orig.reshape((m_test,1))\n",
        "  \n",
        "  U_train = np.sum(train_set_X,axis=1,keepdims=True)/m_train\n",
        "  U_test = np.sum(test_set_X,axis=1,keepdims=True)/m_test\n",
        "  \n",
        "  train_set_X = train_set_X - U_train\n",
        "  test_set_X = test_set_X - U_test\n",
        "  \n",
        "  sigma_train = np.sqrt(np.sum(np.square(train_set_X),axis=1,keepdims=True)/m_train)\n",
        "  sigma_test = np.sqrt(np.sum(np.square(test_set_X),axis=1,keepdims=True)/m_test)\n",
        "  \n",
        "  train_set_X /= sigma_train\n",
        "  test_set_X /= sigma_test\n",
        "  \n",
        "  return train_set_X,train_set_Y,test_set_X,test_set_Y"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kq51oRKnXP4N",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def Data_Preprocessing_Happy():\n",
        "  \n",
        "  train_set_X_orig,train_set_Y_orig,test_set_X_orig,test_set_Y_orig,classes = Load_Data()\n",
        "  train_set_X,train_set_Y,test_set_X,test_set_Y = Reshape_And_Normalize_Happy(train_set_X_orig,train_set_Y_orig,test_set_X_orig,test_set_Y_orig)\n",
        "  \n",
        "  return train_set_X,train_set_Y,test_set_X,test_set_Y"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "J71Oz9IoXJek",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def Propagation_Happy(input_shape,output_neurons):\n",
        "  \n",
        "  X_input = Input(input_shape)\n",
        "  X = X_input\n",
        "  \n",
        "  if(output_neurons > 1):\n",
        "    activation_type = 'softmax'\n",
        "  elif(output_neurons == 1):\n",
        "    activation_type = 'sigmoid'\n",
        "  \n",
        "  X = BatchNormalization(axis=1,name='BatchNormalization')(X)\n",
        "  X = Dense(output_neurons,use_bias=True,activation=activation_type,kernel_initializer='glorot_uniform', bias_initializer='zeros',name='FullyConnected',kernel_regularizer=regularizers.l2(0.01))(X)\n",
        "  \n",
        "  model = keras.models.Model(inputs=X_input,outputs=X,name=\"ANN_Keras\")\n",
        "  \n",
        "  return model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pbCzFdKuW2N3",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def Caller_Function_Happy():\n",
        "  \n",
        "  train_set_X,train_set_Y,test_set_X,test_set_Y = Data_Preprocessing_Happy()\n",
        "  epochs = 10\n",
        "  mini_batch_size = 50\n",
        "  output_neurons = 1\n",
        "  \n",
        "  if(output_neurons > 1):\n",
        "    train_set_Y = np.eye(output_neurons)[train_set_Y.T][0]\n",
        "    test_set_Y = np.eye(output_neurons)[test_set_Y.T][0]\n",
        "    loss_val = 'categorical_crossentropy'\n",
        "  elif(output_neurons == 1):\n",
        "    loss_val = 'binary_crossentropy'\n",
        "    \n",
        "  input_shape = [train_set_X.shape[1]]\n",
        " \n",
        "  model = Propagation_Happy(input_shape,output_neurons)\n",
        "  model.compile(loss=loss_val,optimizer='adam',metrics=['accuracy'])\n",
        "  print(\"\\n---------------------------------------------Happy House Start-------------------------------------------\\n\")\n",
        "  model.fit(x=train_set_X[0:50,:],y=train_set_Y[0:50,:],epochs=epochs,batch_size=mini_batch_size)\n",
        "  \n",
        "  return model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZGncYtH7BMpe",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def Caller_Function():\n",
        "  \n",
        "  model_happy = Caller_Function_Happy()\n",
        "  train_X1,train_X2,train_Y,database = Data_Preprocessing()\n",
        "  epochs = 5\n",
        "  mini_batch_size = 2\n",
        "  output_neurons = 128\n",
        "  \n",
        "  shape = train_X1.shape\n",
        "  shape = (shape[1],shape[2],shape[3])\n",
        " \n",
        "  channel_list = [[96,128,16,32,32,64],[96,128,32,64,64,64],[128,256,32,64,16,32],[96,192,32,64,128,256],[160,256,64,128,32,64],[96,128,64,128,192,256],[192,512,128,256,512,512]]\n",
        "  model = Propagation(shape,output_neurons,channel_list)\n",
        "  opt = Adam(lr=0.5)\n",
        "  model.compile(loss='mean_squared_error',optimizer=opt,metrics=['accuracy'])\n",
        "  print(\"\\n---------------------------------------------Siamese Start-------------------------------------------\\n\")\n",
        "  model.fit(x=[train_X1,train_X2],y=train_Y,epochs=epochs,batch_size=mini_batch_size)\n",
        "  \n",
        "  return model_happy,model,database,train_X1\n",
        "\n",
        "  #model.summary()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KTN4S_ArR4Be",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 672
        },
        "outputId": "df3c5fd2-ac79-4726-eb52-73f0fc90d668"
      },
      "source": [
        "model_happy,model,database,train_X1 = Caller_Function()"
      ],
      "execution_count": 147,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "---------------------------------------------Happy House Start-------------------------------------------\n",
            "\n",
            "Epoch 1/10\n",
            "50/50 [==============================] - 27s 536ms/step - loss: 0.7147 - acc: 0.5800\n",
            "Epoch 2/10\n",
            "50/50 [==============================] - 0s 775us/step - loss: 0.6406 - acc: 0.7000\n",
            "Epoch 3/10\n",
            "50/50 [==============================] - 0s 791us/step - loss: 0.5814 - acc: 0.6800\n",
            "Epoch 4/10\n",
            "50/50 [==============================] - 0s 703us/step - loss: 0.5343 - acc: 0.7200\n",
            "Epoch 5/10\n",
            "50/50 [==============================] - 0s 746us/step - loss: 0.4937 - acc: 0.7200\n",
            "Epoch 6/10\n",
            "50/50 [==============================] - 0s 722us/step - loss: 0.4578 - acc: 0.7400\n",
            "Epoch 7/10\n",
            "50/50 [==============================] - 0s 887us/step - loss: 0.4263 - acc: 0.8600\n",
            "Epoch 8/10\n",
            "50/50 [==============================] - 0s 752us/step - loss: 0.3988 - acc: 0.8600\n",
            "Epoch 9/10\n",
            "50/50 [==============================] - 0s 839us/step - loss: 0.3738 - acc: 0.9200\n",
            "Epoch 10/10\n",
            "50/50 [==============================] - 0s 681us/step - loss: 0.3504 - acc: 0.9400\n",
            "\n",
            "---------------------------------------------Siamese Start-------------------------------------------\n",
            "\n",
            "Epoch 1/5\n",
            "20/20 [==============================] - 29s 1s/step - loss: 0.3082 - acc: 0.6500\n",
            "Epoch 2/5\n",
            "20/20 [==============================] - 2s 81ms/step - loss: 0.2607 - acc: 0.5500\n",
            "Epoch 3/5\n",
            "20/20 [==============================] - 2s 80ms/step - loss: 0.3125 - acc: 0.6500\n",
            "Epoch 4/5\n",
            "20/20 [==============================] - 2s 84ms/step - loss: 0.3174 - acc: 0.6500\n",
            "Epoch 5/5\n",
            "20/20 [==============================] - 2s 81ms/step - loss: 0.2740 - acc: 0.6500\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5u6ka-eqNcHU",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def evaluate(model_happy,model,database,train_X1):\n",
        "  print('')\n",
        "  name = 'arnaud'\n",
        "  model_type = \"verification\"\n",
        "  if(model_type == 'verification'):\n",
        "    image_path = database[name]\n",
        "    anchor = Read_Image(image_path)\n",
        "    i = 10\n",
        "    person = np.zeros((1,64,64,3))\n",
        "    person[0,:,:,:] = train_X1[i,:,:,:]\n",
        "    out = model.predict(x=[anchor,person])\n",
        "    if(out[0][0] >= 0.4):\n",
        "      print('Same',out)\n",
        "      img = person.reshape((person.shape[0],-1))\n",
        "      ans = np.ceil(model_happy.predict(img))\n",
        "      print('Allowed' if ans else 'Rejected')\n",
        "    else:\n",
        "      print('Different',out)\n",
        "  \n",
        "  else:\n",
        "    mx = -100.0\n",
        "    i = 5\n",
        "    person = np.zeros((1,64,64,3))\n",
        "    person[0,:,:,:] = train_X1[i,:,:,:]\n",
        "    \n",
        "    for names in database:\n",
        "      image_path = database[names]\n",
        "      anchor = Read_Image(image_path)\n",
        "      out = model.predict(x=[anchor,person])\n",
        "      if(out[0][0] > mx):\n",
        "        mx = out[0][0]\n",
        "        person_name = names\n",
        "\n",
        "    if(mx >= 0.4):\n",
        "      print('Hello '+person_name)\n",
        "      img = person.reshape((person.shape[0],-1))\n",
        "      ans = np.ceil(model_happy.predict(img))\n",
        "      print('Allowed' if ans else 'Rejected')\n",
        "    else:\n",
        "      print('Not in Database')\n",
        "      "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KF7Tuq1nNfzA",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        },
        "outputId": "237203f0-5576-43aa-a3ca-f7494b032aba"
      },
      "source": [
        "evaluate(model_happy,model,database,train_X1)"
      ],
      "execution_count": 153,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "Different [[0.37951922]]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pMW7HEkKgy6l",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "fafe25ec-c144-4c8a-e86c-3680fc659472"
      },
      "source": [
        "model_happy.summary()\n",
        "model.summary()"
      ],
      "execution_count": 137,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_31 (InputLayer)        (None, 12288)             0         \n",
            "_________________________________________________________________\n",
            "BatchNormalization (BatchNor (None, 12288)             49152     \n",
            "_________________________________________________________________\n",
            "FullyConnected (Dense)       (None, 1)                 12289     \n",
            "=================================================================\n",
            "Total params: 61,441\n",
            "Trainable params: 36,865\n",
            "Non-trainable params: 24,576\n",
            "_________________________________________________________________\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_32 (InputLayer)           (None, 64, 64, 3)    0                                            \n",
            "__________________________________________________________________________________________________\n",
            "input_33 (InputLayer)           (None, 64, 64, 3)    0                                            \n",
            "__________________________________________________________________________________________________\n",
            "zero_padding2d_243 (ZeroPadding (None, 70, 70, 3)    0           input_32[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "zero_padding2d_244 (ZeroPadding (None, 70, 70, 3)    0           input_33[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "Normal_Block_Conv2D_X (Conv2D)  (None, 68, 68, 64)   1792        zero_padding2d_243[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "Normal_Block_Conv2D_Y (Conv2D)  (None, 68, 68, 64)   1792        zero_padding2d_244[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "Normal_Block_BatchNorm_X (Batch (None, 68, 68, 64)   256         Normal_Block_Conv2D_X[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "Normal_Block_BatchNorm_Y (Batch (None, 68, 68, 64)   256         Normal_Block_Conv2D_Y[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "activation_691 (Activation)     (None, 68, 68, 64)   0           Normal_Block_BatchNorm_X[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "activation_692 (Activation)     (None, 68, 68, 64)   0           Normal_Block_BatchNorm_Y[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "Normal_Block_MaxPool1_X (MaxPoo (None, 34, 34, 64)   0           activation_691[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "Normal_Block_MaxPool1_Y (MaxPoo (None, 34, 34, 64)   0           activation_692[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "Normal_Block_AveragePool_X (Ave (None, 17, 17, 64)   0           Normal_Block_MaxPool1_X[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "Normal_Block_AveragePool_Y (Ave (None, 17, 17, 64)   0           Normal_Block_MaxPool1_Y[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "flatten_19 (Flatten)            (None, 18496)        0           Normal_Block_AveragePool_X[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "flatten_20 (Flatten)            (None, 18496)        0           Normal_Block_AveragePool_Y[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "Normal_Layer_FullyConnected_X ( (None, 128)          2367616     flatten_19[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "Normal_Layer_FullyConnected_Y ( (None, 128)          2367616     flatten_20[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "subtract_10 (Subtract)          (None, 128)          0           Normal_Layer_FullyConnected_X[0][\n",
            "                                                                 Normal_Layer_FullyConnected_Y[0][\n",
            "__________________________________________________________________________________________________\n",
            "dense_7 (Dense)                 (None, 50)           6450        subtract_10[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "Siamese_Last_Layer (Dense)      (None, 1)            51          dense_7[0][0]                    \n",
            "==================================================================================================\n",
            "Total params: 4,745,829\n",
            "Trainable params: 4,745,573\n",
            "Non-trainable params: 256\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "akUs2wIfioTk",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}