RestNet.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "anaconda-cloud": {},
    "kernelspec": {
      "display_name": "Python [conda env:TF-1_3]",
      "language": "python",
      "name": "conda-env-TF-1_3-py"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 2
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython2",
      "version": "2.7.13"
    },
    "colab": {
      "name": "RestNet.ipynb",
      "provenance": [],
      "include_colab_link": true
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/Kushagratandon12/Machine-Learning-Projects/blob/master/RestNet.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "collapsed": true,
        "id": "iSBT6MnBkyqm"
      },
      "source": [
        "# Training and Evaluating ResNet Model\n",
        "*by Marvin Bertin*\n",
        "<img src=\"../../images/keras-tensorflow-logo.jpg\" width=\"400\">"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6dHsDPnpkyqt"
      },
      "source": [
        "# Import"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "collapsed": true,
        "id": "hTZvgks1kyqu"
      },
      "source": [
        "from __future__ import absolute_import\n",
        "from __future__ import division\n",
        "from __future__ import print_function\n",
        "\n",
        "import tensorflow as tf\n",
        "import numpy as np\n",
        "import matplotlib\n",
        "import matplotlib.pyplot as plt\n",
        "import os"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "collapsed": true,
        "id": "NNcGM60Fkyqv"
      },
      "source": [
        "models = tf.contrib.keras.models\n",
        "layers = tf.contrib.keras.layers\n",
        "initializers = tf.contrib.keras.initializers\n",
        "regularizers = tf.contrib.keras.regularizers\n",
        "losses = tf.contrib.keras.losses\n",
        "optimizers = tf.contrib.keras.optimizers \n",
        "metrics = tf.contrib.keras.metrics\n",
        "preprocessing_image = tf.contrib.keras.preprocessing.image"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DWt4YnFykyqw"
      },
      "source": [
        "# ResNet Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "collapsed": true,
        "id": "jRFND9yfkyqw"
      },
      "source": [
        "def residual_block(input_tensor, filters, stage, reg=0.0, use_shortcuts=True):\n",
        "\n",
        "    bn_name = 'bn' + str(stage)\n",
        "    conv_name = 'conv' + str(stage)\n",
        "    relu_name = 'relu' + str(stage)\n",
        "    merge_name = 'merge' + str(stage)\n",
        "\n",
        "    # 1x1 conv\n",
        "    # batchnorm-relu-conv\n",
        "    # from input_filters to bottleneck_filters\n",
        "    if stage>1: # first activation is just after conv1\n",
        "        x = layers.BatchNormalization(name=bn_name+'a')(input_tensor)\n",
        "        x = layers.Activation('relu', name=relu_name+'a')(x)\n",
        "    else:\n",
        "        x = input_tensor\n",
        "\n",
        "    x = layers.Convolution2D(\n",
        "            filters[0], (1,1),\n",
        "            kernel_regularizer=regularizers.l2(reg),\n",
        "            use_bias=False,\n",
        "            name=conv_name+'a'\n",
        "        )(x)\n",
        "\n",
        "    # 3x3 conv\n",
        "    # batchnorm-relu-conv\n",
        "    # from bottleneck_filters to bottleneck_filters\n",
        "    x = layers.BatchNormalization(name=bn_name+'b')(x)\n",
        "    x = layers.Activation('relu', name=relu_name+'b')(x)\n",
        "    x = layers.Convolution2D(\n",
        "            filters[1], (3,3),\n",
        "            padding='same',\n",
        "            kernel_regularizer=regularizers.l2(reg),\n",
        "            use_bias = False,\n",
        "            name=conv_name+'b'\n",
        "        )(x)\n",
        "\n",
        "    # 1x1 conv\n",
        "    # batchnorm-relu-conv\n",
        "    # from bottleneck_filters  to input_filters\n",
        "    x = layers.BatchNormalization(name=bn_name+'c')(x)\n",
        "    x = layers.Activation('relu', name=relu_name+'c')(x)\n",
        "    x = layers.Convolution2D(\n",
        "            filters[2], (1,1),\n",
        "            kernel_regularizer=regularizers.l2(reg),\n",
        "            name=conv_name+'c'\n",
        "        )(x)\n",
        "\n",
        "    # merge output with input layer (residual connection)\n",
        "    if use_shortcuts:\n",
        "        x = layers.add([x, input_tensor], name=merge_name)\n",
        "\n",
        "    return x"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "collapsed": true,
        "id": "RxjqrlM5kyqx"
      },
      "source": [
        "def ResNetPreAct(input_shape=(32,32,3), nb_classes=5, num_stages=5,\n",
        "                 use_final_conv=False, reg=0.0):\n",
        "\n",
        "\n",
        "    # Input\n",
        "    img_input = layers.Input(shape=input_shape)\n",
        "\n",
        "    #### Input stream ####\n",
        "    # conv-BN-relu-pool\n",
        "    x = layers.Convolution2D(\n",
        "            128, (3,3), strides=(2, 2),\n",
        "            padding='same',\n",
        "            kernel_regularizer=regularizers.l2(reg),\n",
        "            use_bias=False,\n",
        "            name='conv0'\n",
        "        )(img_input)\n",
        "    x = layers.BatchNormalization(name='bn0')(x)\n",
        "    x = layers.Activation('relu', name='relu0')(x)\n",
        "#     x = layers.MaxPooling2D((3, 3), strides=(2, 2), padding='same', name='pool0')(x)\n",
        "\n",
        "    #### Residual Blocks ####\n",
        "    # 1x1 conv: batchnorm-relu-conv\n",
        "    # FxF conv: batchnorm-relu-conv\n",
        "    # 1x1 conv: batchnorm-relu-conv\n",
        "    for stage in range(1,num_stages+1):\n",
        "        x = residual_block(x, [32,32,128], stage=stage, reg=reg)\n",
        "\n",
        "\n",
        "    #### Output stream ####\n",
        "    # BN-relu-(conv)-avgPool-softmax\n",
        "    x = layers.BatchNormalization(name='bnF')(x)\n",
        "    x = layers.Activation('relu', name='reluF')(x)\n",
        "\n",
        "    # Optional final conv layer\n",
        "    if use_final_conv:\n",
        "        x = layers.Convolution2D(\n",
        "                64, (3,3),\n",
        "                padding='same',\n",
        "                kernel_regularizer=regularizers.l2(reg),\n",
        "                name='convF'\n",
        "            )(x)\n",
        "    \n",
        "    pool_size = input_shape[0] / 2\n",
        "    x = layers.AveragePooling2D((pool_size,pool_size),name='avg_pool')(x)\n",
        "\n",
        "    x = layers.Flatten(name='flat')(x)\n",
        "    x = layers.Dense(nb_classes, activation='softmax', name='fc10')(x)\n",
        "\n",
        "    return models.Model(img_input, x, name='rnpa')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2nrcjuZZkyqz"
      },
      "source": [
        "# Compile Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "collapsed": true,
        "id": "_mxS6iyKkyqz"
      },
      "source": [
        "def compile_model(model):\n",
        "    \n",
        "    # loss\n",
        "    loss = losses.categorical_crossentropy\n",
        "    \n",
        "    # optimizer\n",
        "    optimizer = optimizers.Adam(lr=0.0001)\n",
        "    \n",
        "    # metrics\n",
        "    metric = [metrics.categorical_accuracy, metrics.top_k_categorical_accuracy]\n",
        "    \n",
        "    # compile model with loss, optimizer, and evaluation metrics\n",
        "    model.compile(optimizer, loss, metric)\n",
        "    \n",
        "    return model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-Mdatufmkyq0"
      },
      "source": [
        "# Image Preprocessing And Augmentation"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "collapsed": true,
        "id": "sstSkYgRkyq1"
      },
      "source": [
        "train_datagen = preprocessing_image.ImageDataGenerator(\n",
        "    rescale=1./255,\n",
        "    shear_range=0.1,\n",
        "    zoom_range=0.1,\n",
        "    horizontal_flip=True)\n",
        "\n",
        "test_datagen = preprocessing_image.ImageDataGenerator(rescale=1./255)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SRmPxQtckyq1",
        "outputId": "611f2bcf-f7c6-459f-df8e-786eb35950ce"
      },
      "source": [
        "BASE_DIR = \"/Users/marvinbertin/Desktop/tmp\"\n",
        "\n",
        "train_generator = train_datagen.flow_from_directory(\n",
        "    os.path.join(BASE_DIR, \"flower_dataset/train\"),\n",
        "    target_size=(32, 32),\n",
        "    batch_size=32,\n",
        "    class_mode='categorical')\n",
        "\n",
        "validation_generator = test_datagen.flow_from_directory(\n",
        "    os.path.join(BASE_DIR, \"flower_dataset/validation\"),\n",
        "    target_size=(32, 32),\n",
        "    batch_size=32,\n",
        "    class_mode='categorical')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 2939 images belonging to 5 classes.\n",
            "Found 731 images belonging to 5 classes.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "collapsed": true,
        "id": "_818l4pmkyq4"
      },
      "source": [
        "model = ResNetPreAct(input_shape=(32, 32, 3), nb_classes=5, num_stages=5,\n",
        "                     use_final_conv=False, reg=0.005)\n",
        "\n",
        "model = compile_model(model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MnkV9C-qkyq4"
      },
      "source": [
        "# Train Model on Flower Dataset"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lWW3qzFPkyq5",
        "outputId": "14722136-969d-4693-a2b2-12ad448407a9"
      },
      "source": [
        "history = model.fit_generator(\n",
        "    train_generator,\n",
        "    steps_per_epoch=100,\n",
        "    epochs=10,\n",
        "    validation_data=validation_generator,\n",
        "    validation_steps=20)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "100/100 [==============================] - 97s - loss: 4.7355 - categorical_accuracy: 0.3716 - top_k_categorical_accuracy: 1.0000 - val_loss: 4.7847 - val_categorical_accuracy: 0.2453 - val_top_k_categorical_accuracy: 1.0000\n",
            "Epoch 2/10\n",
            "100/100 [==============================] - 94s - loss: 4.2881 - categorical_accuracy: 0.5327 - top_k_categorical_accuracy: 1.0000 - val_loss: 4.6325 - val_categorical_accuracy: 0.2472 - val_top_k_categorical_accuracy: 1.0000\n",
            "Epoch 3/10\n",
            "100/100 [==============================] - 92s - loss: 3.9952 - categorical_accuracy: 0.5626 - top_k_categorical_accuracy: 1.0000 - val_loss: 4.5625 - val_categorical_accuracy: 0.2441 - val_top_k_categorical_accuracy: 1.0000\n",
            "Epoch 4/10\n",
            "100/100 [==============================] - 93s - loss: 3.7385 - categorical_accuracy: 0.5889 - top_k_categorical_accuracy: 1.0000 - val_loss: 4.4722 - val_categorical_accuracy: 0.2315 - val_top_k_categorical_accuracy: 1.0000\n",
            "Epoch 5/10\n",
            "100/100 [==============================] - 99s - loss: 3.5150 - categorical_accuracy: 0.6122 - top_k_categorical_accuracy: 1.0000 - val_loss: 4.0694 - val_categorical_accuracy: 0.3118 - val_top_k_categorical_accuracy: 1.0000\n",
            "Epoch 6/10\n",
            "100/100 [==============================] - 113s - loss: 3.3142 - categorical_accuracy: 0.6223 - top_k_categorical_accuracy: 1.0000 - val_loss: 3.5542 - val_categorical_accuracy: 0.4142 - val_top_k_categorical_accuracy: 1.0000\n",
            "Epoch 7/10\n",
            "100/100 [==============================] - 90s - loss: 3.1440 - categorical_accuracy: 0.6408 - top_k_categorical_accuracy: 1.0000 - val_loss: 3.2635 - val_categorical_accuracy: 0.5354 - val_top_k_categorical_accuracy: 1.0000\n",
            "Epoch 8/10\n",
            "100/100 [==============================] - 87s - loss: 2.9911 - categorical_accuracy: 0.6518 - top_k_categorical_accuracy: 1.0000 - val_loss: 2.9651 - val_categorical_accuracy: 0.6409 - val_top_k_categorical_accuracy: 1.0000\n",
            "Epoch 9/10\n",
            "100/100 [==============================] - 100s - loss: 2.8415 - categorical_accuracy: 0.6620 - top_k_categorical_accuracy: 1.0000 - val_loss: 2.8743 - val_categorical_accuracy: 0.6079 - val_top_k_categorical_accuracy: 1.0000\n",
            "Epoch 10/10\n",
            "100/100 [==============================] - 119s - loss: 2.7194 - categorical_accuracy: 0.6677 - top_k_categorical_accuracy: 1.0000 - val_loss: 2.7313 - val_categorical_accuracy: 0.6646 - val_top_k_categorical_accuracy: 1.0000\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UZk8pdCvkyq5"
      },
      "source": [
        "# Plot Accuracy And Loss Over Time"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "collapsed": true,
        "id": "_rM71pSHkyq6"
      },
      "source": [
        "def plot_accuracy_and_loss(history):\n",
        "    plt.figure(1, figsize= (15, 10))\n",
        "    \n",
        "    # plot train and test accuracy\n",
        "    plt.subplot(221)\n",
        "    plt.plot(history.history['categorical_accuracy'])\n",
        "    plt.plot(history.history['val_categorical_accuracy'])\n",
        "    plt.title('SqueezeNet accuracy')\n",
        "    plt.ylabel('accuracy')\n",
        "    plt.xlabel('epoch')\n",
        "    plt.legend(['train', 'test'], loc='upper left')\n",
        "\n",
        "    # plot train and test loss\n",
        "    plt.subplot(222)\n",
        "    plt.plot(history.history['loss'])\n",
        "    plt.plot(history.history['val_loss'])\n",
        "    plt.title('SqueezeNet loss')\n",
        "    plt.ylabel('loss')\n",
        "    plt.xlabel('epoch')\n",
        "    plt.legend(['train', 'test'], loc='upper right')\n",
        "    \n",
        "    plt.show()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "luxU3Z0ekyq6",
        "outputId": "2a05994e-c2ed-44d7-a4b4-9d69cfd3c1d0"
      },
      "source": [
        "plot_accuracy_and_loss(history)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3sAAAE0CAYAAACVXCpzAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nT0REREREJBtSsSciIiIiIpIN/T/xhicZa+PEqgAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x1202cfad0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IJQ_4V3hkyq7"
      },
      "source": [
        "# Save Model Weights And Configuration"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "collapsed": true,
        "id": "t8Ud52Q6kyq7"
      },
      "source": [
        "# save model architecture\n",
        "model_json = model.to_json()\n",
        "open('resnet_model.json', 'w').write(model_json)\n",
        "\n",
        "# save model's learned weights\n",
        "model.save_weights('image_classifier_resnet.h5', overwrite=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NJMoSQdOkyq8"
      },
      "source": [
        "## Next Lesson\n",
        "### Xception: Depthwise Separable Convolutions\n",
        "\n",
        "<img src=\"../../images/divider.png\" width=\"100\">"
      ]
    }
  ]
}