Patch AAM .ipynb

{
 "cells": [
  {
   "metadata": {
    "trusted": true,
    "collapsed": false,
    "editable": true,
    "deletable": true
   },
   "cell_type": "code",
   "source": "%matplotlib inline\nimport numpy as np\nfrom pathlib import Path\n\nimport menpo.io as mio\nfrom menpo.feature import fast_dsift\nfrom menpo.transform import AlignmentSimilarity, scale_about_centre\n\nfrom menpodetect import load_dlib_frontal_face_detector\nfrom menpowidgets import visualize_images, visualize_fitting_results\n\nfrom menpofit.aam import LucasKanadeAAMFitter, PatchAAM\n\n# Load the DLIB detector which is fairly strong\n# and has a very low number of false positives\ndetector = load_dlib_frontal_face_detector()",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": true
   },
   "cell_type": "code",
   "source": "face_indices = [4, 5, 6, 7, 8, 9]",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": false,
    "editable": true,
    "deletable": true
   },
   "cell_type": "code",
   "source": "# Loading our images\nimages_folder = Path('/tmp/dataset+landmarks+template');\nimages = mio.import_images(images_folder)",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": false,
    "editable": true,
    "deletable": true
   },
   "cell_type": "code",
   "source": "visualize_images(images)",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": true,
    "editable": true,
    "deletable": true
   },
   "cell_type": "code",
   "source": "# Just a short helper method to crop the images\n# and turn to greyscale - again to save memory\ndef load_image(i):\n    i, tr = i.crop_to_landmarks_proportion(0.3, return_transform=True)\n    if i.n_channels == 3:\n        i = i.as_greyscale()\n    i = i.as_unmasked()\n    i.landmarks['LJSON'] = i.landmarks['LJSON'].add_label('face', face_indices)\n    return i",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": false,
    "editable": true,
    "deletable": true
   },
   "cell_type": "code",
   "source": "training_images = images.map(load_image)",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": false
   },
   "cell_type": "code",
   "source": "visualize_images(training_images)",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "scrolled": true,
    "collapsed": false,
    "editable": true,
    "deletable": true
   },
   "cell_type": "code",
   "source": "aam = PatchAAM(\n    training_images,\n    group='LJSON',\n    verbose=True,\n    holistic_features=fast_dsift, \n    diagonal=200, \n    patch_shape=[(19, 19), (17, 17), (15, 15)],\n    scales=(0.25, 0.5, 1.0)\n)\n\nreference_face_points = aam.reference_shape.get_label('face')\nreference_bb = reference_face_points.bounding_box()\n\naam_fitter = LucasKanadeAAMFitter(aam, n_shape=[3, 5, 10], \n                                  n_appearance=0.8)",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": true
   },
   "cell_type": "code",
   "source": "pretend_test_image = images[10]",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": false,
    "editable": true,
    "deletable": true
   },
   "cell_type": "code",
   "source": "face_bounding_box = detector(pretend_test_image)[0]\ninitial_bb = AlignmentSimilarity(reference_bb,\n                                 face_bounding_box).apply(aam.reference_shape)\ninitial_bb = scale_about_centre(initial_bb, 0.7).apply(initial_bb)\nresult = aam_fitter.fit_from_shape(pretend_test_image, \n                                   initial_bb, \n                                   max_iters=20)",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": false,
    "editable": true,
    "deletable": true
   },
   "cell_type": "code",
   "source": "result.view_widget()",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": true
   },
   "cell_type": "code",
   "source": "mio.export_landmark_file(result.final_shape, '/tmp/result.ljson')",
   "execution_count": null,
   "outputs": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "name": "python3",
   "display_name": "Python 3",
   "language": "python"
  },
  "language_info": {
   "name": "python",
   "mimetype": "text/x-python",
   "nbconvert_exporter": "python",
   "file_extension": ".py",
   "version": "3.5.3",
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "pygments_lexer": "ipython3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}