dev/notebooks/auto_examples/applications/plot_face_recognition.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "\n# Faces recognition example using eigenfaces and SVMs\n\nThe dataset used in this example is a preprocessed excerpt of the\n\"Labeled Faces in the Wild\", aka LFW_:\n\n  http://vis-www.cs.umass.edu/lfw/lfw-funneled.tgz (233MB)\n\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "collapsed": false
      },
      "outputs": [],
      "source": [
        "from time import time\nimport matplotlib.pyplot as plt\n\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.model_selection import RandomizedSearchCV\nfrom sklearn.datasets import fetch_lfw_people\nfrom sklearn.metrics import classification_report\nfrom sklearn.metrics import ConfusionMatrixDisplay\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn.decomposition import PCA\nfrom sklearn.svm import SVC\nfrom scipy.stats import loguniform"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "Download the data, if not already on disk and load it as numpy arrays\n\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "collapsed": false
      },
      "outputs": [],
      "source": [
        "lfw_people = fetch_lfw_people(min_faces_per_person=70, resize=0.4)\n\n# introspect the images arrays to find the shapes (for plotting)\nn_samples, h, w = lfw_people.images.shape\n\n# for machine learning we use the 2 data directly (as relative pixel\n# positions info is ignored by this model)\nX = lfw_people.data\nn_features = X.shape[1]\n\n# the label to predict is the id of the person\ny = lfw_people.target\ntarget_names = lfw_people.target_names\nn_classes = target_names.shape[0]\n\nprint(\"Total dataset size:\")\nprint(\"n_samples: %d\" % n_samples)\nprint(\"n_features: %d\" % n_features)\nprint(\"n_classes: %d\" % n_classes)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "Split into a training set and a test and keep 25% of the data for testing.\n\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "collapsed": false
      },
      "outputs": [],
      "source": [
        "X_train, X_test, y_train, y_test = train_test_split(\n    X, y, test_size=0.25, random_state=42\n)\n\nscaler = StandardScaler()\nX_train = scaler.fit_transform(X_train)\nX_test = scaler.transform(X_test)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "Compute a PCA (eigenfaces) on the face dataset (treated as unlabeled\ndataset): unsupervised feature extraction / dimensionality reduction\n\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "collapsed": false
      },
      "outputs": [],
      "source": [
        "n_components = 150\n\nprint(\n    \"Extracting the top %d eigenfaces from %d faces\" % (n_components, X_train.shape[0])\n)\nt0 = time()\npca = PCA(n_components=n_components, svd_solver=\"randomized\", whiten=True).fit(X_train)\nprint(\"done in %0.3fs\" % (time() - t0))\n\neigenfaces = pca.components_.reshape((n_components, h, w))\n\nprint(\"Projecting the input data on the eigenfaces orthonormal basis\")\nt0 = time()\nX_train_pca = pca.transform(X_train)\nX_test_pca = pca.transform(X_test)\nprint(\"done in %0.3fs\" % (time() - t0))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "Train a SVM classification model\n\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "collapsed": false
      },
      "outputs": [],
      "source": [
        "print(\"Fitting the classifier to the training set\")\nt0 = time()\nparam_grid = {\n    \"C\": loguniform(1e3, 1e5),\n    \"gamma\": loguniform(1e-4, 1e-1),\n}\nclf = RandomizedSearchCV(\n    SVC(kernel=\"rbf\", class_weight=\"balanced\"), param_grid, n_iter=10\n)\nclf = clf.fit(X_train_pca, y_train)\nprint(\"done in %0.3fs\" % (time() - t0))\nprint(\"Best estimator found by grid search:\")\nprint(clf.best_estimator_)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "Quantitative evaluation of the model quality on the test set\n\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "collapsed": false
      },
      "outputs": [],
      "source": [
        "print(\"Predicting people's names on the test set\")\nt0 = time()\ny_pred = clf.predict(X_test_pca)\nprint(\"done in %0.3fs\" % (time() - t0))\n\nprint(classification_report(y_test, y_pred, target_names=target_names))\nConfusionMatrixDisplay.from_estimator(\n    clf, X_test_pca, y_test, display_labels=target_names, xticks_rotation=\"vertical\"\n)\nplt.tight_layout()\nplt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "Qualitative evaluation of the predictions using matplotlib\n\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "collapsed": false
      },
      "outputs": [],
      "source": [
        "def plot_gallery(images, titles, h, w, n_row=3, n_col=4):\n    \"\"\"Helper function to plot a gallery of portraits\"\"\"\n    plt.figure(figsize=(1.8 * n_col, 2.4 * n_row))\n    plt.subplots_adjust(bottom=0, left=0.01, right=0.99, top=0.90, hspace=0.35)\n    for i in range(n_row * n_col):\n        plt.subplot(n_row, n_col, i + 1)\n        plt.imshow(images[i].reshape((h, w)), cmap=plt.cm.gray)\n        plt.title(titles[i], size=12)\n        plt.xticks(())\n        plt.yticks(())"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "plot the result of the prediction on a portion of the test set\n\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "collapsed": false
      },
      "outputs": [],
      "source": [
        "def title(y_pred, y_test, target_names, i):\n    pred_name = target_names[y_pred[i]].rsplit(\" \", 1)[-1]\n    true_name = target_names[y_test[i]].rsplit(\" \", 1)[-1]\n    return \"predicted: %s\\ntrue:      %s\" % (pred_name, true_name)\n\n\nprediction_titles = [\n    title(y_pred, y_test, target_names, i) for i in range(y_pred.shape[0])\n]\n\nplot_gallery(X_test, prediction_titles, h, w)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "plot the gallery of the most significative eigenfaces\n\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "collapsed": false
      },
      "outputs": [],
      "source": [
        "eigenface_titles = [\"eigenface %d\" % i for i in range(eigenfaces.shape[0])]\nplot_gallery(eigenfaces, eigenface_titles, h, w)\n\nplt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "Face recognition problem would be much more effectively solved by training\nconvolutional neural networks but this family of models is outside of the scope of\nthe scikit-learn library. Interested readers should instead try to use pytorch or\ntensorflow to implement such models.\n\n"
      ]
    }
  ],
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.9.16"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}