data_drift_detection_on_cifar10_dataset.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "data-drift-detection-on-cifar10-dataset.ipynb",
      "provenance": [],
      "authorship_tag": "ABX9TyObjZ8j5/gixzo+VY9d/Ijf",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/menon92/DL-Sneak-Peek/blob/master/data_drift_detection_on_cifar10_dataset.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "84SuAnGlf91X",
        "outputId": "28b2048b-0416-4bdd-ab81-c8686395fb4a"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[K     |████████████████████████████████| 329 kB 4.2 MB/s \n",
            "\u001b[K     |████████████████████████████████| 4.8 MB 46.3 MB/s \n",
            "\u001b[K     |████████████████████████████████| 3.8 MB 32.8 MB/s \n",
            "\u001b[K     |████████████████████████████████| 462 kB 46.9 MB/s \n",
            "\u001b[K     |████████████████████████████████| 67 kB 5.6 MB/s \n",
            "\u001b[K     |████████████████████████████████| 895 kB 48.3 MB/s \n",
            "\u001b[K     |████████████████████████████████| 596 kB 49.9 MB/s \n",
            "\u001b[K     |████████████████████████████████| 6.6 MB 34.0 MB/s \n",
            "\u001b[?25h"
          ]
        }
      ],
      "source": [
        "!pip install -q alibi-detect"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import os\n",
        "import tensorflow as tf\n",
        "from alibi_detect.cd import KSDrift\n",
        "from alibi_detect.ad import ModelDistillation\n",
        "\n",
        "from alibi_detect.models.tensorflow.resnet import scale_by_instance\n",
        "from alibi_detect.utils.fetching import fetch_tf_model, fetch_detector\n",
        "from alibi_detect.utils.tensorflow.prediction import predict_batch\n",
        "from alibi_detect.utils.saving import save_detector\n",
        "from alibi_detect.datasets import fetch_cifar10c, corruption_types_cifar10c\n",
        "\n",
        "# source: https://docs.seldon.io/projects/alibi-detect/en/latest/examples/cd_distillation_cifar10.html#Dataset"
      ],
      "metadata": {
        "id": "CBfTe714gEjF"
      },
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "(X_train, y_train), (X_test, y_test) = tf.keras.datasets.cifar10.load_data()\n",
        "X_train = X_train.astype('float32') / 255\n",
        "X_train = scale_by_instance(X_train)\n",
        "y_train = y_train.astype('int64').reshape(-1,)\n",
        "X_test = X_test.astype('float32') / 255\n",
        "y_test = y_test.astype('int64').reshape(-1,)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "k39xkEjHgNO3",
        "outputId": "d8739557-0e24-42b0-dd89-6d61e1615b0b"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz\n",
            "170500096/170498071 [==============================] - 11s 0us/step\n",
            "170508288/170498071 [==============================] - 11s 0us/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "corruptions = corruption_types_cifar10c()\n",
        "print(corruptions)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9JjRoS8igq7K",
        "outputId": "7d5ae132-4093-4a99-f3bb-10fa6a1c9e75"
      },
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "['brightness', 'contrast', 'defocus_blur', 'elastic_transform', 'fog', 'frost', 'gaussian_blur', 'gaussian_noise', 'glass_blur', 'impulse_noise', 'jpeg_compression', 'motion_blur', 'pixelate', 'saturate', 'shot_noise', 'snow', 'spatter', 'speckle_noise', 'zoom_blur']\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "corruption = ['gaussian_noise', 'motion_blur', 'brightness', 'pixelate']\n",
        "X_corr, y_corr = fetch_cifar10c(corruption=corruption, severity=5, return_X_y=True)\n",
        "X_corr = X_corr.astype('float32') / 255"
      ],
      "metadata": {
        "id": "1tc0AoESguTD"
      },
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "X_c = []\n",
        "n_corr = len(corruption)\n",
        "n_test = X_test.shape[0]\n",
        "for i in range(n_corr):\n",
        "    X_c.append(X_corr[i * n_test:(i + 1) * n_test])"
      ],
      "metadata": {
        "id": "C1BgiCsdg6j1"
      },
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "i = 1\n",
        "\n",
        "n_test = X_test.shape[0]\n",
        "plt.title('Original')\n",
        "plt.axis('off')\n",
        "plt.imshow(X_test[i])\n",
        "plt.show()\n",
        "for _ in range(len(corruption)):\n",
        "    plt.title(corruption[_])\n",
        "    plt.axis('off')\n",
        "    plt.imshow(X_corr[n_test * _+ i])\n",
        "    plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "8PyzUTwDhH9g",
        "outputId": "3c8bf6c7-7a19-428b-edac-c066762742bb"
      },
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "dataset = 'cifar10'\n",
        "model = 'resnet32'\n",
        "clf = fetch_tf_model(dataset, model)\n",
        "acc = clf.evaluate(scale_by_instance(X_test), y_test, batch_size=128, verbose=0)[1]\n",
        "print('Test set accuracy:')\n",
        "print('Original {:.4f}'.format(acc))\n",
        "clf_accuracy = {'original': acc}\n",
        "for _ in range(len(corruption)):\n",
        "    acc = clf.evaluate(scale_by_instance(X_c[_]), y_test, batch_size=128, verbose=0)[1]\n",
        "    clf_accuracy[corruption[_]] = acc\n",
        "    print('{} {:.4f}'.format(corruption[_], acc))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "MdAZkO1VhT5p",
        "outputId": "040f9017-6050-4d99-98b2-56cbaeb73ed2"
      },
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/seldon-models/alibi-detect/classifier/cifar10/resnet32/model.h5\n",
            "4300800/4293464 [==============================] - 0s 0us/step\n",
            "4308992/4293464 [==============================] - 0s 0us/step\n",
            "Test set accuracy:\n",
            "Original 0.9278\n",
            "gaussian_noise 0.2208\n",
            "motion_blur 0.6339\n",
            "brightness 0.8913\n",
            "pixelate 0.3666\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        ""
      ],
      "metadata": {
        "id": "YbyWW_amhgdE"
      },
      "execution_count": 8,
      "outputs": []
    }
  ]
}