3D Shape Estimation.ipynb

{
 "cells": [
  {
   "metadata": {
    "trusted": true,
    "collapsed": false
   },
   "cell_type": "code",
   "source": "import menpo3d.io as m3dio\nimport menpo.io as mio\nimport numpy as np\nimport mayavi.mlab as mlab\nimport cv2\n\nfrom menpo.image import Image\n\ntemplate = m3dio.import_mesh('/home/pts08/Dropbox/james/template.obj')\ntemplate_landmarks = mio.import_landmark_file('/home/pts08/Dropbox/james/template.ljson').lms",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": false
   },
   "cell_type": "code",
   "source": "%matplotlib qt\n\ntemplate.view_landmarks()",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": true
   },
   "cell_type": "code",
   "source": "image = mio.import_image('/vol/atlas/databases/lfpw/trainset/image_0002.png')\nimage_landmarks = image.landmarks[None].lms",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": false
   },
   "cell_type": "code",
   "source": "%matplotlib inline\n\nimage.view_widget()",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": false
   },
   "cell_type": "code",
   "source": "rows = image.shape[0]\ncols = image.shape[1]\nmax_d = max(rows, cols)\ncamera_matrix = np.array([[max_d, 0,     rows / 2.0],\n                          [0,     max_d, cols / 2.0],\n                          [0,     0,      1.0]])\ndistortion_coeffs = np.zeros(4)\n\nconverged, r_vec, t_vec = cv2.solvePnP(template_landmarks.points, \n                                       image_landmarks.points[:, ::-1], \n                                       camera_matrix, \n                                       distortion_coeffs)\n\nrotation_matrix = cv2.Rodrigues(r_vec)[0]\nprint(rotation_matrix)",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": false
   },
   "cell_type": "code",
   "source": "img = np.array(image.as_PILImage())\n\nproject_matrix = np.hstack([rotation_matrix, t_vec])\nkp = camera_matrix.dot(project_matrix)\n\n# reproject object points - check validity of found projection matrix\nfor p in template_landmarks.points:\n    X = np.append(p, [1])\n    opt_p = kp.dot(X)\n    opt_p_img = tuple((opt_p[:2] / opt_p[-1]).astype(np.int))\n\n    cv2.circle(img, opt_p_img, 1, (0, 0, 255))",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": false
   },
   "cell_type": "code",
   "source": "%matplotlib inline\n\nImage.init_from_rolled_channels(img).view()",
   "execution_count": null,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true,
    "collapsed": true
   },
   "cell_type": "code",
   "source": "",
   "execution_count": null,
   "outputs": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "name": "python2",
   "display_name": "Python 2",
   "language": "python"
  },
  "language_info": {
   "mimetype": "text/x-python",
   "nbconvert_exporter": "python",
   "name": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.10",
   "file_extension": ".py",
   "codemirror_mode": {
    "version": 2,
    "name": "ipython"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}