Denoising_AutoEncoder.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Denoising_AutoEncoder.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "emUA0zskPTaW"
      },
      "source": [
        "from tensorflow.keras.layers import Dense, Conv2D, Conv2DTranspose, LeakyReLU, Input, Activation, Flatten, Reshape\n",
        "from tensorflow.keras.models import Model\n",
        "from tensorflow.keras.datasets.mnist import load_data\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "from tensorflow.keras.optimizers import Adam\n",
        "from tensorflow.keras.utils import plot_model\n",
        "import matplotlib"
      ],
      "execution_count": 51,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pZFSFmHZnvjC"
      },
      "source": [
        "## Building the encoder"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uizjd0wRCsZ2"
      },
      "source": [
        "latent_dim = 20\n",
        "feature_map = 32"
      ],
      "execution_count": 149,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QJADGUCfpa8D"
      },
      "source": [
        "encoder_input = Input(shape=(28,28,1))\n",
        "encoder_layer1 = Conv2D(feature_map,(3,3))(encoder_input)\n",
        "encoder_layer2 =  LeakyReLU()(encoder_layer1)\n",
        "encoder_flatten = Flatten()(encoder_layer2)\n",
        "encoder_layer3 = Dense(latent_dim)(encoder_flatten)"
      ],
      "execution_count": 150,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Sx4YbwoPqoi6"
      },
      "source": [
        "encoder = Model(encoder_input,encoder_layer3)"
      ],
      "execution_count": 151,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "WTRcak32q1my",
        "outputId": "4bdef2bf-b8f7-46b3-dfda-8fdc5d98b11c"
      },
      "source": [
        "encoder.summary()"
      ],
      "execution_count": 152,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model_19\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_15 (InputLayer)        [(None, 28, 28, 1)]       0         \n",
            "_________________________________________________________________\n",
            "conv2d_7 (Conv2D)            (None, 26, 26, 32)        320       \n",
            "_________________________________________________________________\n",
            "leaky_re_lu_14 (LeakyReLU)   (None, 26, 26, 32)        0         \n",
            "_________________________________________________________________\n",
            "flatten_7 (Flatten)          (None, 21632)             0         \n",
            "_________________________________________________________________\n",
            "dense_19 (Dense)             (None, 20)                432660    \n",
            "=================================================================\n",
            "Total params: 432,980\n",
            "Trainable params: 432,980\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 466
        },
        "id": "c14OWXr5q4Dq",
        "outputId": "2b742402-0df5-4fd6-ec3b-11c133a73913"
      },
      "source": [
        "plot_model(encoder)"
      ],
      "execution_count": 153,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<IPython.core.display.Image object>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 153
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nwaOl9QKvb1o"
      },
      "source": [
        "from tensorflow.keras import backend as K"
      ],
      "execution_count": 154,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Vp1e79sVvc1Q"
      },
      "source": [
        "volumeSize = K.int_shape(encoder_layer2)"
      ],
      "execution_count": 155,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ZHYaFIkXv1id",
        "outputId": "6da6828f-2952-4608-c29a-f33d3b58fa51"
      },
      "source": [
        "volumeSize[1:] "
      ],
      "execution_count": 156,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(26, 26, 32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 156
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BRezUX8Mrmod"
      },
      "source": [
        "## Building the decoder"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xfxaJsEJrFOP"
      },
      "source": [
        "decoder_input = Input(shape=(latent_dim,))\n",
        "\n",
        "decoder_layer1 = Dense(np.prod(volumeSize[1:]))(decoder_input) #Adding the no of neurons that equals to the no of neurons obtained after flattening\n",
        "decoder_input_Reshaped = Reshape((volumeSize[1],volumeSize[2],volumeSize[3]))(decoder_layer1) #Resizing into a matrix for the next layer\n",
        "\n",
        "decoder_layer_cnn = Conv2DTranspose(feature_map,(3,3))(decoder_input_Reshaped)\n",
        "decoder_layer_actv = LeakyReLU()(decoder_layer_cnn)\n",
        "decoder_layer_output = Dense(1)(decoder_layer_actv) #since one depth channel"
      ],
      "execution_count": 157,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RqD8sZc00gTZ"
      },
      "source": [
        "decoder = Model(decoder_input,decoder_layer_output)"
      ],
      "execution_count": 158,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iJi0FG1l0ndj",
        "outputId": "e34f2003-738c-41ba-adc5-1c27619f35ca"
      },
      "source": [
        "decoder.summary()"
      ],
      "execution_count": 159,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model_20\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_16 (InputLayer)        [(None, 20)]              0         \n",
            "_________________________________________________________________\n",
            "dense_20 (Dense)             (None, 21632)             454272    \n",
            "_________________________________________________________________\n",
            "reshape_7 (Reshape)          (None, 26, 26, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_transpose_8 (Conv2DTr (None, 28, 28, 32)        9248      \n",
            "_________________________________________________________________\n",
            "leaky_re_lu_15 (LeakyReLU)   (None, 28, 28, 32)        0         \n",
            "_________________________________________________________________\n",
            "dense_21 (Dense)             (None, 28, 28, 1)         33        \n",
            "=================================================================\n",
            "Total params: 463,553\n",
            "Trainable params: 463,553\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 564
        },
        "id": "WGFgrL1S0o59",
        "outputId": "7e1356e0-be0d-4928-9696-c6f044dce347"
      },
      "source": [
        "plot_model(decoder)"
      ],
      "execution_count": 160,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<IPython.core.display.Image object>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 160
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uvODT3PN8ZyD"
      },
      "source": [
        "autoencoder = Model(encoder_input,decoder(encoder(encoder_input)))"
      ],
      "execution_count": 161,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Otw-Cj5j8z7-",
        "outputId": "eb5fd90d-74e8-40a4-9a76-03af75f9cc05"
      },
      "source": [
        "autoencoder.summary()"
      ],
      "execution_count": 162,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model_21\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_15 (InputLayer)        [(None, 28, 28, 1)]       0         \n",
            "_________________________________________________________________\n",
            "model_19 (Functional)        (None, 20)                432980    \n",
            "_________________________________________________________________\n",
            "model_20 (Functional)        (None, 28, 28, 1)         463553    \n",
            "=================================================================\n",
            "Total params: 896,533\n",
            "Trainable params: 896,533\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "snrPbF028TI1"
      },
      "source": [
        "## Preparing Data with Noise"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "G1ByZoqt0_3p"
      },
      "source": [
        "(x_train_orig,y_train),(x_test_orig,y_test) = load_data()"
      ],
      "execution_count": 163,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mx-ApB7-9hYi"
      },
      "source": [
        "#Normalizing\n",
        "x_train_orig = x_train_orig.astype('float32')/255\n",
        "x_test_orig = x_test_orig.astype('float32')/255"
      ],
      "execution_count": 164,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_LBL4NDM9spr"
      },
      "source": [
        "#Adding noise by sampling from a random normal distribution\n",
        "trainNoise = np.random.normal(loc=0.5, scale=0.5, size=x_train_orig.shape)\n",
        "testNoise = np.random.normal(loc=0.5, scale=0.5, size=x_test_orig.shape)\n",
        "x_train = np.clip(x_train_orig + trainNoise, 0, 1)\n",
        "x_test = np.clip(x_test_orig + testNoise, 0, 1)"
      ],
      "execution_count": 165,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 269
        },
        "id": "Ziez_CiqCDaA",
        "outputId": "4e190e54-ead0-4216-a43d-582f847c6e05"
      },
      "source": [
        "#Noisy input\n",
        "num_images_to_show = 5\n",
        "for im_ind in range(num_images_to_show):\n",
        "    plot_ind = im_ind*2 + 1\n",
        "    rand_ind = np.random.randint(low=0, high=x_train.shape[0])    \n",
        "    matplotlib.pyplot.subplot(num_images_to_show, 2, plot_ind)\n",
        "    matplotlib.pyplot.imshow(x_train[rand_ind, :, :], cmap=\"gray\")\n"
      ],
      "execution_count": 166,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 5 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gSxSTcFT--Bi"
      },
      "source": [
        "#Adding another dimension\n",
        "x_train = np.expand_dims(x_train,axis=-1)\n",
        "x_test = np.expand_dims(x_test,axis=-1)"
      ],
      "execution_count": 167,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YES94c5o_bU3",
        "outputId": "2fefc5dd-b9ce-4259-8e23-9a20b04b2ba8"
      },
      "source": [
        "x_train.shape"
      ],
      "execution_count": 168,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(60000, 28, 28, 1)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 168
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8sP1cG90Iv30"
      },
      "source": [
        "#for the labels\n",
        "x_train_label = np.expand_dims(x_train_orig,axis=-1)\n",
        "x_test_label = np.expand_dims(x_test_orig,axis =-1)"
      ],
      "execution_count": 169,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "NIvVGgHnJb7B",
        "outputId": "ca7d308c-49df-4ea9-961e-c3ed53941501"
      },
      "source": [
        "x_train_label.shape"
      ],
      "execution_count": 170,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(60000, 28, 28, 1)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 170
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TEOZJdN0Ec6k"
      },
      "source": [
        "## Compiling"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1kyyGihhEfck"
      },
      "source": [
        "autoencoder.compile(loss='mse',optimizer=Adam(learning_rate=0.0005))"
      ],
      "execution_count": 173,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_XLesoqq9Fv4"
      },
      "source": [
        "## Training"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JgSSLHac9RM8",
        "outputId": "5c434ddf-5478-46ba-acca-755a702402f0"
      },
      "source": [
        "autoencoder.fit(x_train,x_train_label,batch_size = 256, epochs = 20 , validation_data=(x_test,x_test_label))"
      ],
      "execution_count": 174,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "235/235 [==============================] - 4s 14ms/step - loss: 0.0258 - val_loss: 0.0247\n",
            "Epoch 2/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0243 - val_loss: 0.0236\n",
            "Epoch 3/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0232 - val_loss: 0.0229\n",
            "Epoch 4/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0223 - val_loss: 0.0219\n",
            "Epoch 5/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0216 - val_loss: 0.0215\n",
            "Epoch 6/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0211 - val_loss: 0.0211\n",
            "Epoch 7/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0206 - val_loss: 0.0207\n",
            "Epoch 8/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0202 - val_loss: 0.0204\n",
            "Epoch 9/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0199 - val_loss: 0.0202\n",
            "Epoch 10/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0195 - val_loss: 0.0200\n",
            "Epoch 11/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0193 - val_loss: 0.0199\n",
            "Epoch 12/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0190 - val_loss: 0.0198\n",
            "Epoch 13/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0188 - val_loss: 0.0196\n",
            "Epoch 14/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0186 - val_loss: 0.0195\n",
            "Epoch 15/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0184 - val_loss: 0.0194\n",
            "Epoch 16/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0182 - val_loss: 0.0192\n",
            "Epoch 17/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0180 - val_loss: 0.0192\n",
            "Epoch 18/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0179 - val_loss: 0.0191\n",
            "Epoch 19/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0177 - val_loss: 0.0191\n",
            "Epoch 20/20\n",
            "235/235 [==============================] - 3s 14ms/step - loss: 0.0176 - val_loss: 0.0190\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f4fc44fe490>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 174
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4MfOdvJjALPE"
      },
      "source": [
        "## Predicting"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bbb1ymUDARgK"
      },
      "source": [
        "encoder_prediction = encoder.predict(x_train)"
      ],
      "execution_count": 175,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xMsIY_ZQAcyY"
      },
      "source": [
        "decoder_prediction = decoder.predict(encoder_prediction)"
      ],
      "execution_count": 176,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Pt1kpG3gAgqA"
      },
      "source": [
        "decoder_images =  decoder_prediction.reshape((decoder_prediction.shape[0],28,28))"
      ],
      "execution_count": 177,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 269
        },
        "id": "FFaBE_9tAGRk",
        "outputId": "1a87e4b3-cb1d-433a-e891-9211f7bff9aa"
      },
      "source": [
        "#Plotting the output\n",
        "num_images_to_show = 5\n",
        "for im_ind in range(num_images_to_show):\n",
        "    plot_ind = im_ind*2 + 1\n",
        "    rand_ind = np.random.randint(low=0, high=x_train.shape[0])\n",
        "    matplotlib.pyplot.subplot(num_images_to_show, 2, plot_ind)\n",
        "    matplotlib.pyplot.imshow(x_train_orig[rand_ind, :, :], cmap=\"gray\")\n",
        "    matplotlib.pyplot.subplot(num_images_to_show, 2, plot_ind+1)\n",
        "    matplotlib.pyplot.imshow(decoder_images[rand_ind, :, :], cmap=\"gray\")"
      ],
      "execution_count": 178,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 10 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}