1205_mixed_adversarial_training.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "1205-mixed_adversarial_training.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/parekhakhil/pyImageSearch/blob/main/1205_mixed_adversarial_training.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GV6tBiEtNt0d"
      },
      "source": [
        "![logo_jupyter.png](data:image/png;base64,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)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "c7uPgNc0NyO7"
      },
      "source": [
        "# Mixing normal images and adversarial images when training CNNs\n",
        "\n",
        "\n",
        "### by [PyImageSearch.com](http://www.pyimagesearch.com)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Mpthbjs-N2qJ"
      },
      "source": [
        "## Welcome to **[PyImageSearch Plus](http://pyimg.co/plus)** Jupyter Notebooks!\n",
        "\n",
        "This notebook is associated with the [Mixing normal images and adversarial images when training CNNs](https://www.pyimagesearch.com/2021/03/15/mixing-normal-images-and-adversarial-images-when-training-cnns/) blog post published on 03-15-21.\n",
        "\n",
        "Only the code for the blog post is here. Most codeblocks have a 1:1 relationship with what you find in the blog post with two exceptions: (1) Python classes are not separate files as they are typically organized with PyImageSearch projects, and (2) Command Line Argument parsing is replaced with an `args` dictionary that you can manipulate as needed.\n",
        "\n",
        "We recommend that you execute (press ▶️) the code block-by-block, as-is, before adjusting parameters and `args` inputs. Once you've verified that the code is working, you are welcome to hack with it and learn from manipulating inputs, settings, and parameters. For more information on using Jupyter and Colab, please refer to these resources:\n",
        "\n",
        "*   [Jupyter Notebook User Interface](https://jupyter-notebook.readthedocs.io/en/stable/notebook.html#notebook-user-interface)\n",
        "*   [Overview of Google Colaboratory Features](https://colab.research.google.com/notebooks/basic_features_overview.ipynb)\n",
        "\n",
        "As a reminder, these PyImageSearch Plus Jupyter Notebooks are not for sharing; please refer to the **Copyright** directly below and **Code License Agreement** in the last cell of this notebook. \n",
        "\n",
        "Happy hacking!\n",
        "\n",
        "*Adrian*\n",
        "\n",
        "<hr>\n",
        "\n",
        "***Copyright:*** *The contents of this Jupyter Notebook, unless otherwise indicated, are Copyright 2021 Adrian Rosebrock, PyimageSearch.com. All rights reserved. Content like this is made possible by the time invested by the authors. If you received this Jupyter Notebook and did not purchase it, please consider making future content possible joining PyImageSearch Plus at [http://pyimg.co/plus/](http://pyimg.co/plus) today.*"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0pL5rHGgOAMk"
      },
      "source": [
        "### Download the code zip file"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uEZbwcwepI9V"
      },
      "source": [
        "!wget https://pyimagesearch-code-downloads.s3-us-west-2.amazonaws.com/mixed-adversarial-training/mixed-adversarial-training.zip\n",
        "!unzip -qq mixed-adversarial-training.zip\n",
        "%cd mixed-adversarial-training"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8yYRCnThOQXl"
      },
      "source": [
        "## Blog Post Code"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BcmY4DDJOSIt"
      },
      "source": [
        "### Import Packages"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BlcPho8lOOuF"
      },
      "source": [
        "# import the necessary packages\n",
        "from tensorflow.keras.models import Sequential\n",
        "from tensorflow.keras.layers import BatchNormalization\n",
        "from tensorflow.keras.layers import Conv2D\n",
        "from tensorflow.keras.layers import Activation\n",
        "from tensorflow.keras.layers import Flatten\n",
        "from tensorflow.keras.layers import Dropout\n",
        "from tensorflow.keras.layers import Dense\n",
        "from tensorflow.keras.losses import MSE\n",
        "from tensorflow.keras.optimizers import Adam\n",
        "from tensorflow.keras.utils import to_categorical\n",
        "from tensorflow.keras.datasets import mnist\n",
        "from sklearn.utils import shuffle\n",
        "import matplotlib.pyplot as plt\n",
        "import tensorflow as tf\n",
        "import numpy as np\n",
        "import argparse\n",
        "import cv2"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "uY1ceK63OuFY"
      },
      "source": [
        "### Function to display images in Jupyter Notebooks and Google Colab"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gkPDb7emOuwQ"
      },
      "source": [
        "def plt_imshow(title, image):\n",
        "    # convert the image frame BGR to RGB color space and display it\n",
        "\timage = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n",
        "\tplt.imshow(image)\n",
        "\tplt.title(title)\n",
        "\tplt.grid(False)\n",
        "\tplt.show()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Whv-rpfkWMg2"
      },
      "source": [
        "### Our basic CNN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lurqI0OMWLpr"
      },
      "source": [
        "class SimpleCNN:\n",
        "\t@staticmethod\n",
        "\tdef build(width, height, depth, classes):\n",
        "\t\t# initialize the model along with the input shape\n",
        "\t\tmodel = Sequential()\n",
        "\t\tinputShape = (height, width, depth)\n",
        "\t\tchanDim = -1\n",
        "\n",
        "\t\t# first CONV => RELU => BN layer set\n",
        "\t\tmodel.add(Conv2D(32, (3, 3), strides=(2, 2), padding=\"same\",\n",
        "\t\t\tinput_shape=inputShape))\n",
        "\t\tmodel.add(Activation(\"relu\"))\n",
        "\t\tmodel.add(BatchNormalization(axis=chanDim))\n",
        "\n",
        "\t\t# second CONV => RELU => BN layer set\n",
        "\t\tmodel.add(Conv2D(64, (3, 3), strides=(2, 2), padding=\"same\"))\n",
        "\t\tmodel.add(Activation(\"relu\"))\n",
        "\t\tmodel.add(BatchNormalization(axis=chanDim))\n",
        "\n",
        "\t\t# first (and only) set of FC => RELU layers\n",
        "\t\tmodel.add(Flatten())\n",
        "\t\tmodel.add(Dense(128))\n",
        "\t\tmodel.add(Activation(\"relu\"))\n",
        "\t\tmodel.add(BatchNormalization())\n",
        "\t\tmodel.add(Dropout(0.5))\n",
        "\n",
        "\t\t# softmax classifier\n",
        "\t\tmodel.add(Dense(classes))\n",
        "\t\tmodel.add(Activation(\"softmax\"))\n",
        "\n",
        "\t\t# return the constructed network architecture\n",
        "\t\treturn model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sGB5lf_-WWa_"
      },
      "source": [
        "### Using FGSM to generate adversarial images\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kUOuCBOtWT3L"
      },
      "source": [
        "def generate_image_adversary(model, image, label, eps=2 / 255.0):\n",
        "\t# cast the image\n",
        "\timage = tf.cast(image, tf.float32)\n",
        "\n",
        "\t# record our gradients\n",
        "\twith tf.GradientTape() as tape:\n",
        "\t\t# explicitly indicate that our image should be tacked for\n",
        "\t\t# gradient updates\n",
        "\t\ttape.watch(image)\n",
        "\n",
        "\t\t# use our model to make predictions on the input image and\n",
        "\t\t# then compute the loss\n",
        "\t\tpred = model(image)\n",
        "\t\tloss = MSE(label, pred)\n",
        "\n",
        "\t# calculate the gradients of loss with respect to the image, then\n",
        "\t# compute the sign of the gradient\n",
        "\tgradient = tape.gradient(loss, image)\n",
        "\tsignedGrad = tf.sign(gradient)\n",
        "\n",
        "\t# construct the image adversary\n",
        "\tadversary = (image + (signedGrad * eps)).numpy()\n",
        "\n",
        "\t# return the image adversary to the calling function\n",
        "\treturn adversary"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xi7vMxeuW0Ek"
      },
      "source": [
        "### Updating our data generator to mix normal images with adversarial images on the fly"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vT4mQNiNWxBT"
      },
      "source": [
        "def generate_adversarial_batch(model, total, images, labels, dims,\n",
        "\teps=0.01):\n",
        "\t# unpack the image dimensions into convenience variables\n",
        "\t(h, w, c) = dims\n",
        "\n",
        "\t# we're constructing a data generator here so we need to loop\n",
        "\t# indefinitely\n",
        "\twhile True:\n",
        "\t\t# initialize our perturbed images and labels\n",
        "\t\tperturbImages = []\n",
        "\t\tperturbLabels = []\n",
        "\n",
        "\t\t# randomly sample indexes (without replacement) from the\n",
        "\t\t# input data\n",
        "\t\tidxs = np.random.choice(range(0, len(images)), size=total,\n",
        "\t\t\treplace=False)\n",
        "\n",
        "\t\t# loop over the indexes\n",
        "\t\tfor i in idxs:\n",
        "\t\t\t# grab the current image and label\n",
        "\t\t\timage = images[i]\n",
        "\t\t\tlabel = labels[i]\n",
        "\n",
        "\t\t\t# generate an adversarial image\n",
        "\t\t\tadversary = generate_image_adversary(model,\n",
        "\t\t\t\timage.reshape(1, h, w, c), label, eps=eps)\n",
        "\n",
        "\t\t\t# update our perturbed images and labels lists\n",
        "\t\t\tperturbImages.append(adversary.reshape(h, w, c))\n",
        "\t\t\tperturbLabels.append(label)\n",
        "\n",
        "\t\t# yield the perturbed images and labels\n",
        "\t\tyield (np.array(perturbImages), np.array(perturbLabels))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3xIIiBaIW8_j"
      },
      "source": [
        "def generate_mixed_adverserial_batch(model, total, images, labels,\n",
        "\tdims, eps=0.01, split=0.5):\n",
        "\t# unpack the image dimensions into convenience variables\n",
        "\t(h, w, c) = dims\n",
        "\n",
        "\t# compute the total number of training images to keep along with\n",
        "\t# the number of adversarial images to generate\n",
        "\ttotalNormal = int(total * split)\n",
        "\ttotalAdv = int(total * (1 - split))\n",
        "\n",
        "\t# we're constructing a data generator so we need to loop\n",
        "\t# indefinitely\n",
        "\twhile True:\n",
        "\t\t# randomly sample indexes (without replacement) from the\n",
        "\t\t# input data and then use those indexes to sample our normal\n",
        "\t\t# images and labels\n",
        "\t\tidxs = np.random.choice(range(0, len(images)),\n",
        "\t\t\tsize=totalNormal, replace=False)\n",
        "\t\tmixedImages = images[idxs]\n",
        "\t\tmixedLabels = labels[idxs]\n",
        "\n",
        "\t\t# again, randomly sample indexes from the input data, this\n",
        "\t\t# time to construct our adversarial images\n",
        "\t\tidxs = np.random.choice(range(0, len(images)), size=totalAdv,\n",
        "\t\t\treplace=False)\n",
        "\n",
        "\t\t# loop over the indexes\n",
        "\t\tfor i in idxs:\n",
        "\t\t\t# grab the current image and label, then use that data to\n",
        "\t\t\t# generate the adversarial example\n",
        "\t\t\timage = images[i]\n",
        "\t\t\tlabel = labels[i]\n",
        "\t\t\tadversary = generate_image_adversary(model,\n",
        "\t\t\t\timage.reshape(1, h, w, c), label, eps=eps)\n",
        "\n",
        "\t\t\t# update the mixed images and labels lists\n",
        "\t\t\tmixedImages = np.vstack([mixedImages, adversary])\n",
        "\t\t\tmixedLabels = np.vstack([mixedLabels, label])\n",
        "\n",
        "\t\t# shuffle the images and labels together\n",
        "\t\t(mixedImages, mixedLabels) = shuffle(mixedImages, mixedLabels)\n",
        "\n",
        "\t\t# yield the mixed images and labels to the calling function\n",
        "\t\tyield (mixedImages, mixedLabels)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IlVw-p4IXIsm"
      },
      "source": [
        "### Creating our mixed image and adversarial image training script"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "70RrVrRyXFXB"
      },
      "source": [
        "# load MNIST dataset and scale the pixel values to the range [0, 1]\n",
        "print(\"[INFO] loading MNIST dataset...\")\n",
        "(trainX, trainY), (testX, testY) = mnist.load_data()\n",
        "trainX = trainX / 255.0\n",
        "testX = testX / 255.0\n",
        "\n",
        "# add a channel dimension to the images\n",
        "trainX = np.expand_dims(trainX, axis=-1)\n",
        "testX = np.expand_dims(testX, axis=-1)\n",
        "\n",
        "# one-hot encode our labels\n",
        "trainY = to_categorical(trainY, 10)\n",
        "testY = to_categorical(testY, 10)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yGgCIkY7XWjB"
      },
      "source": [
        "# initialize our optimizer and model\n",
        "print(\"[INFO] compiling model...\")\n",
        "opt = Adam(lr=1e-3)\n",
        "model = SimpleCNN.build(width=28, height=28, depth=1, classes=10)\n",
        "model.compile(loss=\"categorical_crossentropy\", optimizer=opt,\n",
        "\tmetrics=[\"accuracy\"])\n",
        "\n",
        "# train the simple CNN on MNIST\n",
        "print(\"[INFO] training network...\")\n",
        "model.fit(trainX, trainY,\n",
        "\tvalidation_data=(testX, testY),\n",
        "\tbatch_size=64,\n",
        "\tepochs=20,\n",
        "\tverbose=1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "91nvriL4XaUa"
      },
      "source": [
        "# make predictions on the testing set for the model trained on\n",
        "# non-adversarial images\n",
        "(loss, acc) = model.evaluate(x=testX, y=testY, verbose=0)\n",
        "print(\"[INFO] normal testing images:\")\n",
        "print(\"[INFO] loss: {:.4f}, acc: {:.4f}\\n\".format(loss, acc))\n",
        "\n",
        "# generate a set of adversarial from our test set (so we can evaluate\n",
        "# our model performance *before* and *after* mixed adversarial\n",
        "# training)\n",
        "print(\"[INFO] generating adversarial examples with FGSM...\\n\")\n",
        "(advX, advY) = next(generate_adversarial_batch(model, len(testX),\n",
        "\ttestX, testY, (28, 28, 1), eps=0.1))\n",
        "\n",
        "# re-evaluate the model on the adversarial images\n",
        "(loss, acc) = model.evaluate(x=advX, y=advY, verbose=0)\n",
        "print(\"[INFO] adversarial testing images:\")\n",
        "print(\"[INFO] loss: {:.4f}, acc: {:.4f}\\n\".format(loss, acc))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PxA70kd0Xihp"
      },
      "source": [
        "# lower the learning rate and re-compile the model (such that we can\n",
        "# fine-tune it on the mixed batches of normal images and dynamically\n",
        "# generated adversarial images)\n",
        "print(\"[INFO] re-compiling model...\")\n",
        "opt = Adam(lr=1e-4)\n",
        "model.compile(loss=\"categorical_crossentropy\", optimizer=opt,\n",
        "\tmetrics=[\"accuracy\"])\n",
        "\n",
        "# initialize our data generator to create data batches containing\n",
        "# a mix of both *normal* images and *adversarial* images\n",
        "print(\"[INFO] creating mixed data generator...\")\n",
        "dataGen = generate_mixed_adverserial_batch(model, 64,\n",
        "\ttrainX, trainY, (28, 28, 1), eps=0.1, split=0.5)\n",
        "\n",
        "# fine-tune our CNN on the adversarial images\n",
        "print(\"[INFO] fine-tuning network on dynamic mixed data...\")\n",
        "model.fit(\n",
        "\tdataGen,\n",
        "\tsteps_per_epoch=len(trainX) // 64,\n",
        "\tepochs=10,\n",
        "\tverbose=1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xVb_finsXl3k"
      },
      "source": [
        "# now that our model is fine-tuned we should evaluate it on the test\n",
        "# set (i.e., non-adversarial) again to see if performance has degraded\n",
        "(loss, acc) = model.evaluate(x=testX, y=testY, verbose=0)\n",
        "print(\"\")\n",
        "print(\"[INFO] normal testing images *after* fine-tuning:\")\n",
        "print(\"[INFO] loss: {:.4f}, acc: {:.4f}\\n\".format(loss, acc))\n",
        "\n",
        "# do a final evaluation of the model on the adversarial images\n",
        "(loss, acc) = model.evaluate(x=advX, y=advY, verbose=0)\n",
        "print(\"[INFO] adversarial images *after* fine-tuning:\")\n",
        "print(\"[INFO] loss: {:.4f}, acc: {:.4f}\".format(loss, acc))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iIK0y_52aETm"
      },
      "source": [
        "For a detailed walkthrough of the concepts and code, be sure to refer to the full tutorial, [*Mixing normal images and adversarial images when training CNNs*](https://www.pyimagesearch.com/2021/03/15/mixing-normal-images-and-adversarial-images-when-training-cnns/) published on 03-15-21.\n",
        "\n",
        "---\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hwMZ9cZaaLG4"
      },
      "source": [
        "# Code License Agreement\n",
        "```\n",
        "Copyright (c) 2021 PyImageSearch.com\n",
        "\n",
        "SIMPLE VERSION\n",
        "Feel free to use this code for your own projects, whether they are\n",
        "purely educational, for fun, or for profit. THE EXCEPTION BEING if\n",
        "you are developing a course, book, or other educational product.\n",
        "Under *NO CIRCUMSTANCE* may you use this code for your own paid\n",
        "educational or self-promotional ventures without written consent\n",
        "from Adrian Rosebrock and PyImageSearch.com.\n",
        "\n",
        "LONGER, FORMAL VERSION\n",
        "Permission is hereby granted, free of charge, to any person obtaining\n",
        "a copy of this software and associated documentation files\n",
        "(the \"Software\"), to deal in the Software without restriction,\n",
        "including without limitation the rights to use, copy, modify, merge,\n",
        "publish, distribute, sublicense, and/or sell copies of the Software,\n",
        "and to permit persons to whom the Software is furnished to do so,\n",
        "subject to the following conditions:\n",
        "The above copyright notice and this permission notice shall be\n",
        "included in all copies or substantial portions of the Software.\n",
        "Notwithstanding the foregoing, you may not use, copy, modify, merge,\n",
        "publish, distribute, sublicense, create a derivative work, and/or\n",
        "sell copies of the Software in any work that is designed, intended,\n",
        "or marketed for pedagogical or instructional purposes related to\n",
        "programming, coding, application development, or information\n",
        "technology. Permission for such use, copying, modification, and\n",
        "merger, publication, distribution, sub-licensing, creation of\n",
        "derivative works, or sale is expressly withheld.\n",
        "THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND,\n",
        "EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES\n",
        "OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND\n",
        "NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS\n",
        "BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN\n",
        "ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN\n",
        "CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\n",
        "SOFTWARE.\n",
        "```"
      ]
    }
  ]
}