frame_handler.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cv2 as cv\n",
    "\n",
    "import numpy as np\n",
    "\n",
    "%matplotlib tk\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "\n",
    "from PIL import Image\n",
    "import torch\n",
    "\n",
    "import sys\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "load gaze estimator\n",
      "load the pre-trained model:  ckpt/epoch_24_ckpt.pth.tar\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/apramey/final-proj/./utils/model2.py:135: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.\n",
      "  ckpt = torch.load(pre_trained_model_path, map_location=device)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading model\n",
      "Using SAM2\n",
      "Model loaded\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "sys.path.append('./utils')\n",
    "\n",
    "import model2 as gaze_estimation_module\n",
    "gaze_estimation_module.load_model()\n",
    "\n",
    "\n",
    "import model1 as segmentation_module\n",
    "model_path = \"ckpt/sam2.1_hiera_tiny.pt\"\n",
    "SAM2 = \"sam2\"\n",
    "segmentation_module.load_model(None, model_path, SAM2)\n",
    "\n",
    "import image as image_module\n",
    "\n",
    "import mask as mask_module\n",
    "\n",
    "# import yolo as yolo_module"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Only stereo calcbrate one time\n",
    "left_frame0 = 'ProjectFPS24/long_video_left_png/frame_1664.png'\n",
    "right_frame0 = 'ProjectFPS24/long_video_right/frame_1664.jpg'\n",
    "\n",
    "\n",
    "im_left = cv.imread(left_frame0)\n",
    "im_left = cv.cvtColor(im_left, cv.COLOR_BGR2RGB) \n",
    "\n",
    "im_right = cv.imread(right_frame0)\n",
    "im_right = cv.cvtColor(im_right, cv.COLOR_BGR2RGB) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# idea here we need a chessboard to calibrate the camera, world_scaling can be the real size \n",
    "rows = 6\n",
    "cols = 7\n",
    "world_scaling = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_chessboard(im, req=42):\n",
    "    n = 0\n",
    "    req += 1\n",
    "\n",
    "    plt.figure()\n",
    "    plt.imshow(im)\n",
    "    plt.show()\n",
    "    print('Set 42 points to give the chessboard corners')\n",
    "\n",
    "    clicked = plt.ginput(req, timeout=0, show_clicks=True)\n",
    "    if not clicked or len(clicked) < req:\n",
    "        print('Need at least %d lines, you have %d now' % (req, n))\n",
    "\n",
    "\n",
    "    np_clicked = []\n",
    "    for i in clicked[1:]:\n",
    "        np_clicked.append([i])\n",
    "\n",
    "    np_clicked = np.array(np_clicked)\n",
    "    return np_clicked"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# left_corners = get_chessboard(im_left)\n",
    "# np.save(\"left_chessboard\",left_corners)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# right_corners = get_chessboard(im_right)\n",
    "# np.save(\"right_chessboard\", right_corners)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "left_corners=np.load(\"left_chessboard.npy\")\n",
    "left_corners = np.array(left_corners, dtype='float32')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def find_camera_matrix_auto(image):\n",
    "    objp = np.zeros((rows*cols,3), np.float32)\n",
    "    objp[:,:2] = np.mgrid[0:rows,0:cols].T.reshape(-1,2)\n",
    "    objp = objp * world_scaling\n",
    "    width = image.shape[1]\n",
    "    height = image.shape[0]   \n",
    "\n",
    "    img_point = []\n",
    "    points_3d = []\n",
    "    gray = cv.cvtColor(image, cv.COLOR_BGR2GRAY)\n",
    "    criteria = (cv.TERM_CRITERIA_EPS + cv.TERM_CRITERIA_MAX_ITER, 30, 0.001)\n",
    "    ret, corners = cv.findChessboardCorners(gray, (rows,cols), cv.CALIB_CB_NORMALIZE_IMAGE)\n",
    "    if ret == True:\n",
    "        corners2 = cv.cornerSubPix(gray, corners, (11,11), (-1,-1), criteria)\n",
    "        img_point.append(corners2)\n",
    "        points_3d.append(objp)\n",
    "\n",
    "        # cv.drawChessboardCorners(image, (rows,cols), corners2, ret)\n",
    "        # cv.imshow('image', image)\n",
    "        # cv.waitKey(0)\n",
    "        # cv.destroyAllWindows()\n",
    "    else:\n",
    "        print('corners not found')\n",
    "    ret, mtx, dist, rvecs, tvecs = cv.calibrateCamera(points_3d,img_point, (width, height), None, None)\n",
    "    print(ret)\n",
    "    return mtx, dist, img_point, points_3d\n",
    "\n",
    "def find_camera_matrix_manual(image, corners):\n",
    "    objp = np.zeros((rows*cols,3), np.float32)\n",
    "    objp[:,:2] = np.mgrid[0:rows,0:cols].T.reshape(-1,2)\n",
    "    objp = objp * world_scaling\n",
    "    width = image.shape[1]\n",
    "    height = image.shape[0]   \n",
    "\n",
    "    img_point = []\n",
    "    points_3d = []\n",
    "    img_point.append(corners)\n",
    "    points_3d.append(objp)\n",
    "    gray = cv.cvtColor(image, cv.COLOR_BGR2GRAY)\n",
    "\n",
    "    # cv.imshow('image', image)\n",
    "    # cv.waitKey(0)\n",
    "    # cv.destroyAllWindows()\n",
    "\n",
    "\n",
    "    ret, mtx, dist, rvecs, tvecs = cv.calibrateCamera(points_3d,img_point, (width, height), None, None)\n",
    "    print(ret)\n",
    "    return mtx, dist, img_point, points_3d\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.15130642723709312\n",
      "0.41953839100710766\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "mtx_R, dist_R, img_point_R, points_3d_R = find_camera_matrix_auto(im_right)\n",
    "\n",
    "mtx_L, dist_L, img_point_L, points_3d_L =  find_camera_matrix_manual(im_left, left_corners)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# points = []\n",
    "# for i in img_point_R[0]:\n",
    "#     points.append(i[0])\n",
    "# points = np.array(points[:7])\n",
    "# print(points)\n",
    "# plt.figure()\n",
    "# plt.imshow(im_right)\n",
    "# plt.scatter(points[:,0], points[:,1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[-0.22851414  0.21614433  0.01018155 -0.012397    0.15947108]]\n",
      "[[-7.48070944e+00  1.87660564e+02 -2.67514700e-01  9.34490885e-03\n",
      "  -1.89821993e+03]]\n",
      "1920 1080\n",
      "3840 2160\n"
     ]
    }
   ],
   "source": [
    "print(dist_L)\n",
    "print(dist_R)\n",
    "\n",
    "\n",
    "width = im_left.shape[1]\n",
    "height = im_left.shape[0]\n",
    "print(width, height)\n",
    "width = im_right.shape[1]\n",
    "height = im_right.shape[0]\n",
    "print(width, height)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def find_stereo_camera_matrix(mtx_L, dist_L, mtx_R, dist_R, img_point_L, img_point_R, points_3d):\n",
    "    stereocalibration_flags = cv.CALIB_FIX_INTRINSIC\n",
    "    criteria = (cv.TERM_CRITERIA_EPS + cv.TERM_CRITERIA_MAX_ITER, 100, 0.0001)\n",
    "    ret, CM1, dist1, CM2, dist2, R, T, E, F = cv.stereoCalibrate(points_3d, img_point_L, img_point_R, mtx_L, dist_L,\n",
    "                                mtx_R, dist_R, (width, height), criteria = criteria, flags = stereocalibration_flags)\n",
    "    return R, T\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "R,T = find_stereo_camera_matrix(mtx_L, dist_L, mtx_R, dist_R, img_point_L, img_point_R, points_3d_L)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "RT1 = np.concatenate([np.eye(3), [[0],[0],[0]]], axis = -1)\n",
    "P1 = mtx_L @ RT1\n",
    "\n",
    "RT2 = np.concatenate([R, T], axis = -1)\n",
    "P2 = mtx_R @ RT2\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "def find_center(M, shown = False):\n",
    "    U, S, V = np.linalg.svd(M)\n",
    "    if shown:\n",
    "        print(S)\n",
    "        print(\"Camera Center:\", V[-1, :3] / V[-1, -1])\n",
    "    return V[-1, :3] / V[-1, -1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def DLT(P1, P2, point1, point2):\n",
    " \n",
    "    A = [point1[1]*P1[2,:] - P1[1,:],\n",
    "         P1[0,:] - point1[0]*P1[2,:],\n",
    "         point2[1]*P2[2,:] - P2[1,:],\n",
    "         P2[0,:] - point2[0]*P2[2,:]\n",
    "        ]\n",
    "    A = np.array(A).reshape((4,4))\n",
    "    #print('A: ')\n",
    "    #print(A)\n",
    " \n",
    "    B = A.transpose() @ A\n",
    "    from scipy import linalg\n",
    "    U, s, Vh = linalg.svd(B, full_matrices = False)\n",
    " \n",
    "    # print('Triangulated point: ')\n",
    "    # print(Vh[3,0:3]/Vh[3,3])\n",
    "    return Vh[3,0:3]/Vh[3,3]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[-58.8905832  -18.87801782 128.99025931]\n"
     ]
    }
   ],
   "source": [
    "pt = DLT(P1, P2, img_point_L[0][0][0], img_point_R[0][0][0])\n",
    "print(pt)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "pt = DLT(P1, P2, img_point_L[0][1][0], img_point_R[0][1][0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0. 0. 0.]\n",
      "[66.7928762   3.35899459 -0.07095524]\n"
     ]
    }
   ],
   "source": [
    "c1 = find_center(P1)\n",
    "c2 = find_center(P2)\n",
    "\n",
    "print(c1)\n",
    "print(c2)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_input_lines(im, min_lines=4):\n",
    "    n = 0\n",
    "\n",
    "    plt.figure()\n",
    "    plt.imshow(im)\n",
    "    plt.show()\n",
    "    print('Set 4 points to give the plane of table')\n",
    "\n",
    "    clicked = plt.ginput(min_lines, timeout=0, show_clicks=True)\n",
    "    if not clicked or len(clicked) < min_lines:\n",
    "        print('Need at least %d lines, you have %d now' % (min_lines, n))\n",
    "\n",
    "\n",
    "    \n",
    "    pt1 = np.array([clicked[0][0], clicked[0][1], 1])\n",
    "    pt2 = np.array([clicked[1][0], clicked[1][1], 1])\n",
    "    pt3 = np.array([clicked[2][0], clicked[2][1], 1])\n",
    "    pt4 = np.array([clicked[3][0], clicked[3][1], 1])\n",
    "\n",
    "    return [pt1, pt2, pt3, pt4]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# pts_left = (get_input_lines(im_left, 4))\n",
    "# pts_right = (get_input_lines(im_right, 4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# np.save(\"pts_left\", pts_left)\n",
    "# np.save(\"pts_right\", pts_right)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "pts_left = np.load(\"pts_left.npy\")\n",
    "pts_right = np.load(\"pts_right.npy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "42 42 42\n"
     ]
    }
   ],
   "source": [
    "pts_x = []\n",
    "pts_y = []\n",
    "pts_z = []\n",
    "\n",
    "for i in range(img_point_L[0].shape[0]):\n",
    "    pt = DLT(P1, P2, img_point_L[0][i][0], img_point_R[0][i][0])\n",
    "    # pt = pt / pt[2]\n",
    "    pts_x.append(pt[0])\n",
    "    pts_y.append(pt[1])\n",
    "    pts_z.append(pt[2])\n",
    "print(len(pts_x), len(pts_y), len(pts_z))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "pt_table = np.array([DLT(P1, P2, pts_left[0], pts_right[0]), DLT(P1, P2, pts_left[1], pts_right[1]), DLT(P1, P2, pts_left[2], pts_right[2])])\n",
    "\n",
    "v1 = pt_table[0] - pt_table[1]\n",
    "v2 = pt_table[0] - pt_table[2]\n",
    "normal = np.cross(v1, v2)\n",
    "\n",
    "d = -np.dot(normal, pt_table[0])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "fig = plt.figure()\n",
    "\n",
    "ax_3d = fig.add_subplot(111, projection='3d')\n",
    "\n",
    "pts_x_c = pts_x + [c1[0], c2[0]]\n",
    "pts_y_c = pts_y + [c1[1], c2[1]]\n",
    "pts_z_c = pts_z + [c1[2], c2[2]]\n",
    "\n",
    "\n",
    "max_range = max([max(pts_x_c)-min(pts_x_c), max(pts_y_c)-min(pts_y_c), max(pts_z_c)-min(pts_z_c)])\n",
    "mid_x = (max(pts_x_c) + min(pts_x_c)) * 0.5\n",
    "mid_y = (max(pts_y_c) + min(pts_y_c)) * 0.5\n",
    "mid_z = (max(pts_z_c) + min(pts_z_c)) * 0.5\n",
    "ax_3d.view_init(elev=135, azim=0)\n",
    "ax_3d.set_xlim(mid_x - max_range/2, mid_x + max_range/2)\n",
    "ax_3d.set_ylim(mid_y - max_range/2, mid_y + max_range/2)\n",
    "ax_3d.set_zlim(mid_z - max_range/2, mid_z + max_range/2)\n",
    "\n",
    "ax_3d.scatter(pts_x, pts_y, pts_z, c='b')  \n",
    "ax_3d.scatter(c1[0], c1[1], c1[2], c='r')\n",
    "ax_3d.scatter(c2[0], c2[1], c2[2], c='r')\n",
    "ax_3d.scatter(pt_table[0][0], pt_table[0][1], pt_table[0][2], c='g')\n",
    "ax_3d.scatter(pt_table[1][0], pt_table[1][1], pt_table[1][2], c='g')\n",
    "ax_3d.scatter(pt_table[2][0], pt_table[2][1], pt_table[2][2], c='g')\n",
    "X_range = np.linspace(mid_x - max_range/2, mid_x + max_range/2, 10)\n",
    "Y_range = np.linspace(mid_y - max_range/2, mid_y + max_range/2, 10)\n",
    "ax_3d.legend(['Chessboard', 'Camera 1', 'Camera 2', 'Table'])\n",
    "n_x, n_y, n_z = normal\n",
    "X, Y = np.meshgrid(X_range, Y_range)\n",
    "Z = (-d - n_x * X - n_y * Y) / n_z\n",
    "ax_3d.plot_surface(X, Y, Z, alpha=0.5, color='cyan')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "def find_plane_two_vectors(c, d1, d2):\n",
    "\n",
    "    n = np.cross(d1, d2)  \n",
    "    if np.linalg.norm(n) == 0:\n",
    "        raise ValueError(\"Direction vectors are collinear; no unique plane can be defined.\")\n",
    "    \n",
    "    n = n / np.linalg.norm(n)  \n",
    "    d = -np.dot(n, c) \n",
    "\n",
    "    return n, d"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "def backproject(P, x):\n",
    "\n",
    "    x_h = np.array([x[0], x[1], 1])  \n",
    "    M = P[:, :3]  \n",
    "    d = np.linalg.inv(M) @ x_h \n",
    "    d = d / np.linalg.norm(d)  \n",
    "    return d\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "def intersection_line_of_planes(n1, d1, n2, d2):\n",
    "    \"\"\"\n",
    "    Find the intersection line of two planes in 3D.\n",
    "\n",
    "    Parameters:\n",
    "        n1 (numpy.ndarray): Normal vector of plane 1.\n",
    "        d1 (float): Plane 1 constant.\n",
    "        n2 (numpy.ndarray): Normal vector of plane 2.\n",
    "        d2 (float): Plane 2 constant.\n",
    "\n",
    "    Returns:\n",
    "        X0 (numpy.ndarray): A point on the intersection line.\n",
    "        d (numpy.ndarray): Direction vector of the intersection line.\n",
    "    \"\"\"\n",
    "    # Step 1: Direction of the intersection line (cross product of normals)\n",
    "    d =  np.cross(n1, n2)\n",
    "    \n",
    "    # Check if the planes are parallel\n",
    "    if np.linalg.norm(d) == 0:\n",
    "        raise ValueError(\"Planes are parallel, no intersection line.\")\n",
    "    \n",
    "    # Step 2: Find a point on the intersection line\n",
    "    # Solve for a point X0 satisfying both plane equations\n",
    "    A = np.vstack((n1, n2, d))  # Stack normal vectors and direction vector\n",
    "    b = np.array([d1, d2, 0])   # Right-hand side with 0 for d direction\n",
    "    \n",
    "    # Solve the linear system for X0\n",
    "    X0 = - np.linalg.lstsq(A, b, rcond=None)[0]\n",
    "    \n",
    "    return X0, d\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Gaze estimation on all the frames"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1080, 1920, 3)\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.04297274 -0.3647134 ]\n"
     ]
    }
   ],
   "source": [
    "img, small = image_module.load_img(left_frame0)\n",
    "image_module.set_img(small)\n",
    "\n",
    "print(img.shape)\n",
    "# gaze estimation\n",
    "\n",
    "arrow_start_left, arrow_end_left = gaze_estimation_module.find_vector_arrow(img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.figure()\n",
    "plt.imshow(img)\n",
    "plt.scatter(arrow_start_left[0], arrow_start_left[1], c='r')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.41596723 -0.1681314 ]\n"
     ]
    }
   ],
   "source": [
    "img, small = image_module.load_img(right_frame0)\n",
    "\n",
    "arrow_start_right, arrow_end_right = gaze_estimation_module.find_vector_arrow(img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.figure()\n",
    "plt.imshow(img)\n",
    "plt.scatter(arrow_start_right[0], arrow_start_right[1], c='r')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "face_3d = DLT(P1, P2, arrow_start_left, arrow_start_right)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "v_left_1 = backproject(P1, arrow_start_left)\n",
    "v_left_2 = backproject(P1, arrow_end_left)\n",
    "\n",
    "v_right_1 = backproject(P2, arrow_start_right)\n",
    "v_right_2 = backproject(P2, arrow_end_right)\n",
    "\n",
    "n_left, d_left = find_plane_two_vectors(face_3d, v_left_1, v_left_2)\n",
    "n_right, d_right = find_plane_two_vectors(c2, v_right_1, v_right_2)\n",
    "\n",
    "c_, d_line = intersection_line_of_planes(n_left, d_left, n_right, d_right)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "fig = plt.figure()\n",
    "\n",
    "ax_3d = fig.add_subplot(111, projection='3d')\n",
    "\n",
    "pts_x_c = pts_x + [c1[0], c2[0]]\n",
    "pts_y_c = pts_y + [c1[1], c2[1]]\n",
    "pts_z_c = pts_z + [c1[2], c2[2]]\n",
    "\n",
    "\n",
    "max_range = max([max(pts_x_c)-min(pts_x_c), max(pts_y_c)-min(pts_y_c), max(pts_z_c)-min(pts_z_c)])\n",
    "mid_x = (max(pts_x_c) + min(pts_x_c)) * 0.5\n",
    "mid_y = (max(pts_y_c) + min(pts_y_c)) * 0.5\n",
    "mid_z = (max(pts_z_c) + min(pts_z_c)) * 0.5\n",
    "ax_3d.view_init(elev=135, azim=0)\n",
    "ax_3d.set_xlim(mid_x - max_range/2, mid_x + max_range/2)\n",
    "ax_3d.set_ylim(mid_y - max_range/2, mid_y + max_range/2)\n",
    "ax_3d.set_zlim(mid_z - max_range/2, mid_z + max_range/2)\n",
    "\n",
    "ax_3d.scatter(pts_x, pts_y, pts_z, c='b', label='Chessboard')\n",
    "ax_3d.scatter(c1[0], c1[1], c1[2], c='r', label='Camera 1')\n",
    "ax_3d.scatter(c2[0], c2[1], c2[2], c='purple',  label='Camera 2')\n",
    "ax_3d.scatter(face_3d[0], face_3d[1], face_3d[2], c='black', label='Face')\n",
    "ax_3d.scatter(pt_table[0][0], pt_table[0][1], pt_table[0][2], c='g', label='Table')\n",
    "ax_3d.scatter(pt_table[1][0], pt_table[1][1], pt_table[1][2], c='g' )\n",
    "ax_3d.scatter(pt_table[2][0], pt_table[2][1], pt_table[2][2], c='g')\n",
    "X_range = np.linspace(mid_x - max_range/2, mid_x + max_range/2, 10)\n",
    "Y_range = np.linspace(mid_y - max_range/2, mid_y + max_range/2, 10)\n",
    "n_x, n_y, n_z = normal\n",
    "X, Y = np.meshgrid(X_range, Y_range)\n",
    "Z = (-d - n_x * X - n_y * Y) / n_z\n",
    "ax_3d.plot_surface(X, Y, Z, alpha=0.5, color='cyan')\n",
    "t = np.linspace(-200, 200, 100)\n",
    "\n",
    "\n",
    "X = c_[0] + t * d_line[0]\n",
    "Y = c_[1] + t * d_line[1]\n",
    "Z = c_[2] + t * d_line[2]\n",
    "\n",
    "\n",
    "ax_3d.plot(X, Y, Z, c='orange', label='Gaze vector')\n",
    "ax_3d.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "# def get_masks_on_table(frame_path, resize_scale=0.5):\n",
    "#     img, small = image_module.load_img(frame_path, resize_scale)\n",
    "#     image_module.set_img(img)\n",
    "\n",
    "#     # gaze estimation\n",
    "#     arrow_start, arrow_end = gaze_estimation_module.find_vector_arrow(img)\n",
    "\n",
    "#     # yolo implementation (finds table)\n",
    "#     table_img, curr_table = yolo_module.get_table(small)\n",
    "#     reset_table_bbox = torch.tensor([0, 0, curr_table[2]-curr_table[0], curr_table[3]-curr_table[1]])\n",
    "\n",
    "#     #segmentation\n",
    "#     SAM2 = \"sam2\"\n",
    "#     masks = segmentation_module.mask_generate(table_img, SAM2)\n",
    "#     masks_on_table = mask_module.find_object_given_table(reset_table_bbox, masks)\n",
    "\n",
    "#     new_masks = []\n",
    "\n",
    "#     for mask in masks_on_table:\n",
    "#         seg = mask[\"segmentation\"]\n",
    "\n",
    "#         full_size_seg = np.zeros(img.shape[:-1], dtype=bool)\n",
    "#         full_size_seg[int(curr_table[1]):int(curr_table[3])+1,int(curr_table[0]):int(curr_table[2])+1] = seg\n",
    "\n",
    "#         mask['segmentation'] = full_size_seg\n",
    "        \n",
    "\n",
    "#     return masks_on_table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_masks_on_table(frame_path, is_left, resize_scale=0.5):\n",
    "    img, small = image_module.load_img(frame_path, resize_scale)\n",
    "    image_module.set_img(img)\n",
    "\n",
    "    # gaze estimation\n",
    "    arrow_start, arrow_end = gaze_estimation_module.find_vector_arrow(img)\n",
    "\n",
    "    # Get table bounding box from points left\n",
    "    # We have three corners\n",
    "    if is_left:\n",
    "        corner1 = pts_left[0][:2]\n",
    "        corner2 = pts_left[1][:2]\n",
    "        corner3 = pts_left[2][:2]\n",
    "        corner4 = pts_left[3][:2]\n",
    "    else:\n",
    "        corner1 = pts_right[0][:2]\n",
    "        corner2 = pts_right[1][:2]\n",
    "        corner3 = pts_right[2][:2]\n",
    "        corner4 = pts_right[3][:2]\n",
    "\n",
    "    xs = np.array([corner1[1], corner2[1], corner3[1], corner4[1]])\n",
    "    ys = np.array([corner1[0], corner2[0], corner3[0], corner4[0]])\n",
    "    max_x = np.max(xs)\n",
    "    max_y = np.max(ys)\n",
    "\n",
    "    min_x = np.min(xs)\n",
    "    min_y = np.min(ys)\n",
    "\n",
    "\n",
    "    table_bbox = np.array([min_y, min_x,max_y - min_y, max_x - min_x])\n",
    "\n",
    "\n",
    "    #segment the whole table\n",
    "    SAM2 = \"sam2\"\n",
    "    masks = segmentation_module.mask_generate(img, SAM2)\n",
    "    # image_module.show_box(masks[7]['bbox'])\n",
    "    masks_on_table = mask_module.find_object_given_table(table_bbox, masks)\n",
    "\n",
    "\n",
    "    return masks_on_table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.04297274 -0.3647134 ]\n",
      "Generating mask\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.41596723 -0.1681314 ]\n",
      "Generating mask\n"
     ]
    }
   ],
   "source": [
    "left_frame_masks = get_masks_on_table(left_frame0, True)\n",
    "\n",
    "image_module.set_img(image_module.load_img(left_frame0)[0])\n",
    "image_module.show_anns(left_frame_masks)\n",
    "plt.show()\n",
    "\n",
    "right_frame_masks = get_masks_on_table(right_frame0, False)\n",
    "\n",
    "image_module.set_img(image_module.load_img(right_frame0)[0])\n",
    "image_module.show_anns(right_frame_masks)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "def reconstruct_3d(masks_left, masks_right):\n",
    "    pt_mask_right = []\n",
    "    pt_mask_left = []\n",
    "    for mask in masks_left:\n",
    "        y_coords, x_coords = np.where(mask[\"segmentation\"])\n",
    "        x = np.mean(x_coords)\n",
    "        y = np.mean(y_coords)\n",
    "        pt_mask_left.append([x, y])\n",
    "\n",
    "    for mask in masks_right:\n",
    "        y_coords, x_coords = np.where(mask[\"segmentation\"])\n",
    "        x = np.mean(x_coords)\n",
    "        y = np.mean(y_coords)\n",
    "        pt_mask_right.append([x, y])\n",
    "\n",
    "    pt_mask_3d = []\n",
    "    indexs = []\n",
    "    for i in range(len(pt_mask_left)):\n",
    "        min_dis = 1000000\n",
    "        cur_pt = None\n",
    "        index = -1\n",
    "        for j in range(len(pt_mask_right)):\n",
    "            pt = DLT(P1, P2, pt_mask_left[i], pt_mask_right[j])\n",
    "            dist = abs(np.dot(normal, pt ) + d)\n",
    "            if dist < min_dis:\n",
    "                min_dis = dist\n",
    "                cur_pt = pt\n",
    "                index = j\n",
    "            # pt_mask_3d.append(pt)\n",
    "        pt_mask_3d.append(cur_pt)\n",
    "        indexs.append(index)\n",
    "    return np.array(pt_mask_3d), np.array(indexs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "fig = plt.figure()\n",
    "\n",
    "ax_3d = fig.add_subplot(111, projection='3d')\n",
    "\n",
    "pts_x_c = pts_x + [c1[0], c2[0]]\n",
    "pts_y_c = pts_y + [c1[1], c2[1]]\n",
    "pts_z_c = pts_z + [c1[2], c2[2]]\n",
    "\n",
    "\n",
    "max_range = max([max(pts_x_c)-min(pts_x_c), max(pts_y_c)-min(pts_y_c), max(pts_z_c)-min(pts_z_c)])\n",
    "mid_x = (max(pts_x_c) + min(pts_x_c)) * 0.5\n",
    "mid_y = (max(pts_y_c) + min(pts_y_c)) * 0.5\n",
    "mid_z = (max(pts_z_c) + min(pts_z_c)) * 0.5\n",
    "ax_3d.view_init(elev=135, azim=0)\n",
    "ax_3d.set_xlim(mid_x - max_range/2, mid_x + max_range/2)\n",
    "ax_3d.set_ylim(mid_y - max_range/2, mid_y + max_range/2)\n",
    "ax_3d.set_zlim(mid_z - max_range/2, mid_z + max_range/2)\n",
    "\n",
    "ax_3d.scatter(pts_x, pts_y, pts_z, c='b', label='Chessboard')\n",
    "ax_3d.scatter(c1[0], c1[1], c1[2], c='r', label='Camera 1')\n",
    "ax_3d.scatter(c2[0], c2[1], c2[2], c='purple',  label='Camera 2')\n",
    "ax_3d.scatter(face_3d[0], face_3d[1], face_3d[2], c='gold', label='Face')\n",
    "ax_3d.scatter(pt_table[0][0], pt_table[0][1], pt_table[0][2], c='g', label='Table')\n",
    "ax_3d.scatter(pt_table[1][0], pt_table[1][1], pt_table[1][2], c='g' )\n",
    "ax_3d.scatter(pt_table[2][0], pt_table[2][1], pt_table[2][2], c='g')\n",
    "X_range = np.linspace(mid_x - max_range/2, mid_x + max_range/2, 10)\n",
    "Y_range = np.linspace(mid_y - max_range/2, mid_y + max_range/2, 10)\n",
    "n_x, n_y, n_z = normal\n",
    "X, Y = np.meshgrid(X_range, Y_range)\n",
    "Z = (-d - n_x * X - n_y * Y) / n_z\n",
    "ax_3d.plot_surface(X, Y, Z, alpha=0.5, color='cyan')\n",
    "t = np.linspace(-200, 200, 100)\n",
    "\n",
    "\n",
    "X = c_[0] + t * d_line[0]\n",
    "Y = c_[1] + t * d_line[1]\n",
    "Z = c_[2] + t * d_line[2]\n",
    "\n",
    "\n",
    "ax_3d.plot(X, Y, Z, c='orange', label='Gaze vector')\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "################################################################################\n",
    "pointsd, indexs = reconstruct_3d(left_frame_masks, right_frame_masks)\n",
    "\n",
    "ax_3d.scatter(pointsd[:, 0], pointsd[:, 1], pointsd[:,2], c='black', label='Reconstructed points')\n",
    "\n",
    "ax_3d.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "pointsd, indexs = reconstruct_3d(left_frame_masks, right_frame_masks)\n",
    "\n",
    "# ax_3d.scatter(pointsd[:, 0], pointsd[:, 1], pointsd[:,2], c='black', label='Reconstructed points')\n",
    "\n",
    "# ax_3d.legend()\n",
    "# plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "# plt.imshow(image_module.load_img(left_frame0)[0])\n",
    "# plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "# plt.imshow(image_module.load_img(right_frame0)[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "def find_closest_point(points, x0, d):\n",
    "    # x0: intercept point on the line, a vector [x0_x, x0_y, x0_z]\n",
    "    # d: direction vector of the line [d_x, d_y, d_z]\n",
    "    # points: list of points [[x1, y1, z1], [x2, y2, z2], ...]\n",
    "    \n",
    "    closest_point = None\n",
    "    index = -1\n",
    "    min_dist_sq = float('inf')\n",
    "    \n",
    "    dd = d[0]*d[0] + d[1]*d[1] + d[2]*d[2]  # d·d\n",
    "    \n",
    "    for i in range(len(points)):\n",
    "        p = points[i]\n",
    "        px, py, pz = p\n",
    "        t = ((px - x0[0])*d[0] + (py - x0[1])*d[1] + (pz - x0[2])*d[2]) / dd\n",
    "\n",
    "        proj = [x0[0] + t*d[0], x0[1] + t*d[1], x0[2] + t*d[2]]\n",
    "        \n",
    "        dx = p[0] - proj[0]\n",
    "        dy = p[1] - proj[1]\n",
    "        dz = p[2] - proj[2]\n",
    "        dist_sq = dx*dx + dy*dy + dz*dz\n",
    "\n",
    "        if dist_sq < min_dist_sq:\n",
    "            min_dist_sq = dist_sq\n",
    "            closest_point = p\n",
    "            index = i\n",
    "    \n",
    "    return closest_point, index, min_dist_sq\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "closest_point, index, mds = find_closest_point(pointsd, c_, d_line)\n",
    "# pointsd.pop(index)\n",
    "# second_closest_point, index2 = find_closest_point(pointsd, c_, d_line)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'segmentation': array([[False, False, False, ..., False, False, False],\n",
       "        [False, False, False, ..., False, False, False],\n",
       "        [False, False, False, ..., False, False, False],\n",
       "        ...,\n",
       "        [False, False, False, ..., False, False, False],\n",
       "        [False, False, False, ..., False, False, False],\n",
       "        [False, False, False, ..., False, False, False]]),\n",
       " 'area': 1968,\n",
       " 'bbox': [1087.0, 675.0, 69.0, 40.0],\n",
       " 'predicted_iou': 0.9593369364738464,\n",
       " 'point_coords': [[1110.0, 691.875]],\n",
       " 'stability_score': 0.9733668565750122,\n",
       " 'crop_box': [0.0, 0.0, 1920.0, 1080.0]}"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "left_frame_masks[index]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "left_list = os.listdir('ProjectFPS24/long_video_left_png')\n",
    "left_list.sort(key=lambda x: x[6:10])\n",
    "left_list = left_list[1653:-9]\n",
    "\n",
    "right_list = os.listdir('ProjectFPS24/long_video_right')\n",
    "right_list.sort(key=lambda x:x[6:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "detected_faces:  1\n",
      "pred_gaze_np:  [ 0.02944824 -0.15444033]\n",
      "Generating mask\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.41596723 -0.1681314 ]\n",
      "Generating mask\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [ 0.02944824 -0.15444033]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.41596723 -0.1681314 ]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [ 0.02641964 -0.12324595]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.50371486 -0.18448395]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [ 0.02369107 -0.14577402]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4920427  -0.17034237]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.04224747 -0.21444286]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5560435  -0.12618935]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.01626143 -0.18067637]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5436598  -0.14569414]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.10642676 -0.16047259]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.56892383 -0.07619727]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.1117917  -0.17445567]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5900352  -0.20485033]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.07670464 -0.18543068]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.577163   -0.14546685]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.06829792 -0.19566755]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5713657  -0.19517012]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.04720647 -0.27412808]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.44300044 -0.1435989 ]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.04297274 -0.3647134 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5313056 -0.194923 ]\n",
      "5\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/apramey/final-proj/./utils/image.py:26: RuntimeWarning: More than 20 figures have been opened. Figures created through the pyplot interface (`matplotlib.pyplot.figure`) are retained until explicitly closed and may consume too much memory. (To control this warning, see the rcParam `figure.max_open_warning`). Consider using `matplotlib.pyplot.close()`.\n",
      "  plt.figure(figsize=(10,10))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.07007199 -0.36699638]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6271393   0.42653325]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.09720109 -0.35949996]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6579491   0.47391677]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.18221647 -0.23029993]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6075465   0.48065928]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.18132925 -0.20404841]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.59777343  0.35453987]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.18892194 -0.2132721 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5407214  0.3585325]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.20376436 -0.23886147]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5730104   0.31714815]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.34007683 -0.21613519]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5291611   0.80492324]\n",
      "10\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3016677 -0.3150484]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.44870174  0.88875365]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.25097567 -0.3965979 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.47985473  0.9725633 ]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.28471842 -0.214599  ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.54035044  1.0168853 ]\n",
      "10\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.30076817 -0.0665357 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4433837  0.9512627]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3645808  -0.08032985]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.48012877  1.0065992 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.2934338  -0.04644064]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.40260828  0.94868714]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.2658107  -0.09213281]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.34220266  0.9115354 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.26572952 -0.07117391]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5435423  1.0543376]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.27829784 -0.12126824]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.44359118  1.0202947 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.30841383 -0.03321964]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.44034487  0.9130437 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.25810242  0.17645721]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6632897   0.75682294]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.390168   -0.05377564]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5904093  0.9499448]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5186892 -0.2875601]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.451223   0.9859636]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.29747286 -0.15898983]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4334607  0.9568911]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5012324 -0.1891201]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.42436403  0.90222096]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.39571837 -0.01765677]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.36682868  0.8858796 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.36566925 -0.01457144]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.37850752  0.817228  ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.37332973 -0.1963884 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.43187556  0.94612366]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33499435 -0.03628823]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3770281   0.79283977]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33217585 -0.10820812]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.39772168  0.90901244]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.39744407 -0.11206472]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4277247   0.91465104]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.39129984 -0.08465429]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3640641   0.91962904]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.43406466 -0.07420873]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3263691  0.8233265]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.45763573 -0.03149886]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.42223072  0.75961566]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.46502694  0.03637312]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.44508737  0.8903105 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.46319038 -0.05193312]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.42017666  0.8589816 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4384152  -0.00894117]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4564061  0.6864252]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.38452718 -0.04141292]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.50927883  0.75935495]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3985888  -0.01943553]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4443912  0.7400537]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3349083   0.00287514]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.45468572  0.87067866]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.34499827  0.00598009]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.55292445  0.88929874]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.39137003  0.03593405]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.38849178  0.7625854 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.35210004 -0.02516056]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.40074015  0.8072442 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.36768654 -0.01364306]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33438697  0.8219498 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33389467 -0.02571834]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.2876524   0.92912996]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.2880794   0.04557304]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.35851866  0.8733862 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.28044933  0.04225905]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.36064383  0.85887355]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.26615474  0.06618733]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.34689918  0.83976436]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.25712433  0.03992693]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.38937345  0.88922584]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.25221142 -0.01483724]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3864364  0.9709679]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.28652918 -0.11463354]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.41698802  0.78390044]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.29768756 -0.07419787]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.45850706  1.1047611 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3409095  -0.05973228]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.50386906  0.86369604]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5537542   0.15383807]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4678415  0.7111651]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4922763  -0.01883353]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.41227975  0.8074571 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4201337  -0.12311161]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.28879508  0.7847902 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.35154077 -0.06037775]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.32083952  0.82314974]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.35235056 -0.19266744]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.32629508  0.7715365 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3551113  -0.19620664]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.32949787  0.77890515]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3614971  -0.16641702]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3482163  0.6938668]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33529374 -0.12496914]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.43159038  0.84851277]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3237236  -0.13154595]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3170314  1.0075482]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.29408237 -0.12781942]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.43353918  0.8778499 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.28063184 -0.11945134]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.40531635  0.89803714]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.29586336 -0.03091148]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.42946208  0.85895413]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3211936  -0.13700242]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.45312196  0.8227848 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3141944  -0.14674011]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4555275   0.83545333]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.27764326 -0.16197446]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4049433   0.85707474]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.29337618 -0.12914957]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33955657  0.884306  ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.30370674 -0.04229008]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.39841133  0.89192295]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.23732498 -0.01279963]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3778948   0.81743765]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.32036898 -0.0929391 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3122377   0.75317335]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.269347   -0.12330069]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.36891955  0.68252474]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.24746259 -0.11747466]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.36813644  0.87266254]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.24914005 -0.14183268]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3966474   0.83660984]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.25498715 -0.12488995]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.39832556  0.8583628 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.27091077 -0.12374675]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.35921454  0.7706372 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.2681899  -0.13373135]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.29700333  0.85438305]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.27692613 -0.09591314]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.31389874  0.78675884]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.25040495 -0.1288129 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.37784338  0.8046683 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.21308745 -0.09383743]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33075836  0.74156654]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.26753226 -0.08473257]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3319173  0.8321403]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.2916826  -0.14863198]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.39484984  0.8113867 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.26857066 -0.16229017]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33927724  0.7621106 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.2513662  -0.12169808]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.35167545  0.7538997 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.21363717 -0.0948208 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33874056  0.7625169 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.26939255 -0.1229564 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4490803  0.8581348]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.25304985 -0.08654101]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7072957   0.87763685]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.24025486 -0.0565745 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.52366424  0.761625  ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.34025276 -0.05910581]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.48286426  0.8829647 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4183411  -0.03577779]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4832833  0.7325689]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.36914015  0.00924051]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.44502443  0.80012846]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.31188378  0.0342203 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.47586116  0.837964  ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.29428217  0.03258616]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.48958173  0.80323106]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33545604  0.05572023]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.46019876  0.7879926 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.34066886  0.07456414]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.41596037  0.7140156 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.23508939  0.03659471]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4412802  0.8295611]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.2453813   0.03491704]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.38896927  0.82131904]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.28294384  0.0388134 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.39707118  0.7091532 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.20492369  0.06408244]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5006628  0.9644337]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.31050605  0.09596382]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.46573368  0.84023434]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3108777   0.10327343]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.44689414  0.9048371 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.37026814  0.14242986]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5125511  0.7469009]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.31242794  0.04370108]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.58372307  0.92551833]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3097961   0.03802633]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5298981   0.83072567]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.30068108  0.01152104]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5932923  0.8744537]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.30561206  0.01252445]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.52942   0.827919]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.28105116  0.03406086]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.36451107  0.7781017 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3597568   0.04629923]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.426364   0.7723025]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33126017  0.05762048]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.25204968  0.83834434]\n",
      "0\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.34049937  0.0995068 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.35133958  0.61485046]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.30315533 -0.00424118]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.48170155  0.3922046 ]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.29649588 -0.02271621]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.45102674  0.8055779 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.42196196  0.21607724]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4339898   0.85737973]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.41103068  0.17529349]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.35894647  0.6076151 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3801833   0.16859539]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4323755   0.87311816]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.42145422  0.15450779]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.39317632  0.5563965 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3804683   0.15763997]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.52261424  0.8097369 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.392978    0.17547397]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.50523525  0.5472629 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.34533316  0.10745797]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.45867512  0.5811133 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.36598292  0.1352229 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.50271773  0.6304642 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.37814468  0.1375646 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4315737  0.7510947]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.40807313  0.13118406]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5104033   0.78502476]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.42759812  0.1260597 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.46265668  0.5986857 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.41769376  0.10919292]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.51001465  0.5911234 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.49046105  0.11727987]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5926637   0.85215586]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.41106012  0.1613857 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5049735   0.78467166]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.46945855  0.1328248 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4546225  0.9079118]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.42209005  0.0979551 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.46914008  0.74248666]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.47331738  0.1546    ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5407338  1.0148467]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5375109   0.21984138]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5045296  0.8506245]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5249024   0.14637965]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.44199347  1.0194548 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5585966   0.08591486]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.45244014  1.0158055 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.51746225  0.106436  ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4448598  1.0248836]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3759187  -0.02833466]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.38849878  0.98011285]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.46591216  0.00745209]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.47870192  0.789353  ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4522514   0.02127719]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.53699374  0.5591736 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.34909382 -0.01077254]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.48261937  0.89474934]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.34953687 -0.01191962]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4599704  0.8149899]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3899409   0.07687987]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4945404  1.0320929]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.39527124  0.08392616]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.44804147  0.8976578 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.38367873  0.09158222]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.36890623  0.7223249 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.42481193  0.07807675]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.37444016  0.8289566 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.38805416  0.0685123 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4643036  0.976627 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.38003936  0.04213474]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.38664585  1.1306846 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33777922  0.04979545]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.49223852  0.7409492 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3672066   0.04720203]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4821107   0.69424105]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.37761787  0.08981553]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5206593   0.87099916]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.39270443  0.1472968 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4808011  0.7348184]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.39738774  0.07586542]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.482695   0.8327188]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.34514695  0.02249178]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4401092   0.76194954]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33130914  0.02242797]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.51944536  0.8117993 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3560026   0.08322813]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.497551    0.85943425]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.29499266 -0.03373094]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.24376823  0.8349368 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.37321     0.13268742]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.34033012  0.78330076]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.36671066  0.03731702]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.47164562  0.7347187 ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.42299795  0.11399364]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.40440786  0.79482555]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3953348   0.06828527]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.47926626  0.90629363]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.38949955  0.09223376]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.56602407  0.954255  ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.37407044  0.08460517]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.38955694  0.86485   ]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3317562  -0.02725156]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.46134022  0.7867562 ]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.34834445  0.01701202]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.49793085  0.97020555]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.30212593  0.00781299]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4504726  0.7666589]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.38611805  0.14779975]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.50140095  1.0453146 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.24636504  0.02041424]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.58327955  0.7960583 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.344822   0.1429706]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.44856822  0.9418208 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3199434   0.07089763]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5587094  0.9291808]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33165368  0.08664522]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4753883  1.0076064]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3088172   0.09347173]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5836072  0.9781328]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.34494662  0.13426374]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.50207764  0.7955526 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.31129438  0.04815069]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5310569   0.88889295]\n",
      "1\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3280498   0.18747404]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6211249   0.92951876]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.27842265 -0.00791495]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5150469   0.91658384]\n",
      "6\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.38460526  0.20939651]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5798364  0.7783874]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33537713  0.1142736 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.57420796  0.9771827 ]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.31224975  0.15295961]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.588186    0.94484437]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.33443117  0.16348678]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6390602   0.79467976]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.27534088  0.05013639]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6839893   0.50307125]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.41457245  0.04172432]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6390742   0.46089846]\n",
      "9\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5150285  -0.38434306]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6265002   0.49258414]\n",
      "8\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.508598   -0.28684676]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5747574   0.44037113]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.42478982 -0.4444464 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6020694  0.4435188]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4338956 -0.4732875]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5648626   0.24507001]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.40817755 -0.48929456]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.75416416  0.3460172 ]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4274317 -0.5156303]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6728846   0.26674438]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5041812  -0.55675846]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6338013   0.40216315]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4896867 -0.4655989]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7404178   0.37706658]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.37645492 -0.49020252]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.64394665  0.51057744]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3413415 -0.5132859]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7221427   0.47258985]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.32191163 -0.41512814]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.74629664  0.523337  ]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.35039556 -0.39280424]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7389929   0.49335057]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.21135722 -0.21602592]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6237394  0.4315486]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.26309434 -0.25694397]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.72522295  0.32996103]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.2054871  -0.26452285]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7971129   0.47696188]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.29846805 -0.2806032 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.70137596  0.45845896]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3571242  -0.39855966]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7073479   0.51837516]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3181092 -0.3879469]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.68161553  0.4989739 ]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.24827771 -0.2255161 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6786821   0.53554344]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.30296206 -0.08456761]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.73148423  0.51885694]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.31714985 -0.04801948]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.72578555  0.46588555]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.3460704  -0.09530641]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.77896136  0.45391563]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.25588256 -0.06726411]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.754875   0.4586219]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.18322368 -0.18808109]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6437171  0.3902868]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.1605602  -0.19583394]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.64107984  0.39533737]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.16455176 -0.19389297]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.80961883  0.42270246]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.16377139 -0.2862269 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.79793566  1.0891845 ]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.12457998 -0.3136606 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7556742  0.724165 ]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.24225453 -0.370524  ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6965706   0.50234056]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.21170315 -0.28372818]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.71465814  0.53305364]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.25614914 -0.24080606]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6715331  0.5115745]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.14753123 -0.31950015]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7298379  0.5167793]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.14647323 -0.2982744 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.64730823  0.47053462]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.15051645 -0.29931942]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.61844057  0.45694292]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.1990922  -0.30733043]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6173953   0.45366037]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.19053358 -0.32014564]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.64868706  0.38940254]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.20630966 -0.28359607]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.65215826  0.31882986]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.20853062 -0.25891152]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.602661    0.30818477]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.19988257 -0.262843  ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.652457   0.5242455]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.19498442 -0.30302992]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6010271   0.52592534]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.21681948 -0.2523331 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.64729184  0.5741273 ]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.2167904  -0.28487745]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.63482356  0.43810487]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.17852221 -0.33714166]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5501495   0.37976715]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.19740334 -0.33572397]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5913514   0.52334756]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.20871261 -0.3107817 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.63622713  0.44688883]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.23982249 -0.29135734]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7122687   0.38230914]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.21123016 -0.28282186]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.63688517  0.43431836]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.23289357 -0.271567  ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6439946   0.48197374]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.20811632 -0.26823008]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.71627665  0.4501512 ]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.22720635 -0.27949834]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.76796925  0.4655499 ]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.17921735 -0.2676658 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6793902   0.51531553]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.17699417 -0.29125103]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.70499486  0.50934213]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.13508508 -0.21768202]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.76665914  0.573691  ]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.15639247 -0.27965018]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.70945317  0.49207273]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.1671398  -0.26730663]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.65459836  0.44181836]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.18328676 -0.21068   ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7756257  0.3364049]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.20707169 -0.24160925]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7178406  0.4068792]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.21827188 -0.17318933]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.65671676  0.40077063]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.17379619 -0.24336602]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.77387184  0.40639788]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.16816892 -0.28245974]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7682283  0.4503846]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.17349657 -0.2415397 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.65945095  0.51003504]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.19099237 -0.22270995]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5833554  0.5132394]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.15904644 -0.18293677]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6474859   0.37728605]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.1819751  -0.18088074]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.68748695  0.40003476]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.15129988 -0.17899095]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.70354337  0.5750081 ]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.193683   -0.17783615]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.58211505  0.5281912 ]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.16412392 -0.1928395 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6226487  0.5072877]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.17495428 -0.13514522]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7007727  0.3636534]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.14452355 -0.19209717]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.707946    0.55456793]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.21963952 -0.16722502]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6848419   0.52301145]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.21253255 -0.1629222 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7295035   0.40074766]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.19990088 -0.14825304]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.697538    0.43370715]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.16424477 -0.14976226]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6989307   0.39840227]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.03513627 -0.08571979]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7260892  0.546868 ]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.14147301 -0.13199647]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.7909241  0.4387354]\n",
      "11\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.17456949 -0.14956932]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6378031   0.43988925]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.12050156 -0.19017677]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6388719  0.5803289]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.17406654 -0.16593003]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.72600526  0.6362071 ]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.11524659 -0.18244432]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.48755026  0.45761058]\n",
      "3\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.08731303 -0.19299069]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.73838925 -0.21197811]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.06649487 -0.21808974]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6014417   0.21636657]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.07757457 -0.21758494]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.58614814 -0.22434041]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.11299833 -0.20026217]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.619652  -0.2554685]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.06559072 -0.43320337]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5806858  -0.26819673]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [ 0.00695542 -0.33201548]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5944886  -0.31696388]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.05880193 -0.3639713 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.46412146 -0.17127055]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.05094912 -0.2961029 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.6313374  -0.22540577]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.19396926 -0.5336157 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5244476  -0.22427809]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.17673357 -0.4683643 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.47326496 -0.2082062 ]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.1286884  -0.33492133]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.54007375 -0.29539803]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.1798092 -0.2933071]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5426054  -0.28815678]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.20892568 -0.2530117 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5771481  -0.22949225]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.15396526 -0.29201424]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.58140564 -0.17939049]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.13199532 -0.22715858]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5159014  -0.21868868]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.22702041 -0.28362894]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5112824  -0.19201949]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.22591186 -0.30029625]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.63780147 -0.3301718 ]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.19311588 -0.31947878]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.50134844 -0.2646061 ]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.15387888 -0.2030653 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5218438 -0.3668108]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.13959667 -0.1803166 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.51068777 -0.2631252 ]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.1865789  -0.28862175]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5059158  -0.22516207]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.1329481  -0.23365338]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.38860524 -0.25700563]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.08636993 -0.19895782]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5941669  -0.18243824]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.13697556 -0.22413109]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5560899  -0.31460002]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.09209491 -0.21771662]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5787777  -0.31195578]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.11495347 -0.23790868]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.47607073 -0.2945462 ]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.11943705 -0.2049404 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5402149 -0.3073293]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.0044527  -0.16011782]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5250499 -0.3020185]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [ 0.00890318 -0.17939685]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.60508597 -0.29025346]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.01415848 -0.21691972]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.607453   -0.25631604]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.05191348 -0.18144004]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5004365  -0.24707673]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.01338243 -0.17614393]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5198689  -0.22072597]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.00691265 -0.17991124]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.431409   -0.18948758]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [ 0.00845678 -0.14158513]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.52901345 -0.27720052]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.02484727 -0.16555953]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.4677865  -0.27284408]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.04283926 -0.17241625]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5527291  -0.22287653]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.10672327 -0.3510558 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.47195256 -0.19141148]\n",
      "2\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.01804651 -0.14003621]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5183383  -0.20518233]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.00611916 -0.23184459]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.56990916 -0.3740673 ]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [ 0.01710374 -0.14640218]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5922874  -0.28406042]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [ 0.02637828 -0.1976506 ]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.43194318 -0.22972688]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [ 0.04723189 -0.14103998]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.42498413 -0.21447605]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [ 0.03508193 -0.15117598]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.46758404 -0.18085535]\n",
      "5\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [ 0.00932988 -0.13185784]\n",
      "detected_faces:  1\n",
      "pred_gaze_np:  [-0.5281791  -0.23983715]\n",
      "5\n",
      "warning: no detected face\n"
     ]
    },
    {
     "ename": "NameError",
     "evalue": "name 'exit' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[53], line 12\u001b[0m\n\u001b[1;32m      8\u001b[0m     pointsd, indexs \u001b[38;5;241m=\u001b[39m reconstruct_3d(left_frame_masks, right_frame_masks)\n\u001b[1;32m     11\u001b[0m im_left, small_left \u001b[38;5;241m=\u001b[39m image_module\u001b[38;5;241m.\u001b[39mload_img(left_frame)\n\u001b[0;32m---> 12\u001b[0m arrow_start_left, arrow_end_left \u001b[38;5;241m=\u001b[39m \u001b[43mgaze_estimation_module\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mfind_vector_arrow\u001b[49m\u001b[43m(\u001b[49m\u001b[43mim_left\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mcopy\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m     16\u001b[0m im_right, small_right \u001b[38;5;241m=\u001b[39m image_module\u001b[38;5;241m.\u001b[39mload_img(right_frame)\n\u001b[1;32m     17\u001b[0m arrow_start_right, arrow_end_right \u001b[38;5;241m=\u001b[39m gaze_estimation_module\u001b[38;5;241m.\u001b[39mfind_vector_arrow(im_right\u001b[38;5;241m.\u001b[39mcopy())\n",
      "File \u001b[0;32m~/final-proj/./utils/model2.py:158\u001b[0m, in \u001b[0;36mfind_vector_arrow\u001b[0;34m(image)\u001b[0m\n\u001b[1;32m    156\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mlen\u001b[39m(detected_faces) \u001b[38;5;241m==\u001b[39m \u001b[38;5;241m0\u001b[39m:\n\u001b[1;32m    157\u001b[0m     \u001b[38;5;28mprint\u001b[39m(\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mwarning: no detected face\u001b[39m\u001b[38;5;124m'\u001b[39m)\n\u001b[0;32m--> 158\u001b[0m     \u001b[43mexit\u001b[49m(\u001b[38;5;241m0\u001b[39m)\n\u001b[1;32m    159\u001b[0m \u001b[38;5;28mprint\u001b[39m(\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mdetected_faces: \u001b[39m\u001b[38;5;124m'\u001b[39m, \u001b[38;5;28mlen\u001b[39m(detected_faces))\n\u001b[1;32m    160\u001b[0m best_face \u001b[38;5;241m=\u001b[39m find_largest_face(detected_faces)\n",
      "\u001b[0;31mNameError\u001b[0m: name 'exit' is not defined"
     ]
    }
   ],
   "source": [
    "for i in range(len(left_list)):\n",
    "    left_frame = os.path.join('ProjectFPS24/long_video_left_png/', f\"{left_list[i]}\")\n",
    "    right_frame = os.path.join('ProjectFPS24/long_video_right/', f\"{right_list[i]}\")\n",
    "    if i == 0:\n",
    "        left_frame_masks = get_masks_on_table(left_frame, True)\n",
    "        right_frame_masks = get_masks_on_table(right_frame, False)\n",
    "\n",
    "        pointsd, indexs = reconstruct_3d(left_frame_masks, right_frame_masks)\n",
    "\n",
    "\n",
    "    im_left, small_left = image_module.load_img(left_frame)\n",
    "    arrow_start_left, arrow_end_left = gaze_estimation_module.find_vector_arrow(im_left.copy())\n",
    "\n",
    "\n",
    "\n",
    "    im_right, small_right = image_module.load_img(right_frame)\n",
    "    arrow_start_right, arrow_end_right = gaze_estimation_module.find_vector_arrow(im_right.copy())\n",
    "\n",
    "\n",
    "    face_3d = DLT(P1, P2, arrow_start_left, arrow_start_right)\n",
    "\n",
    "    v_left_1 = backproject(P1, arrow_start_left)\n",
    "    v_left_2 = backproject(P1, arrow_end_left)\n",
    "\n",
    "    v_right_1 = backproject(P2, arrow_start_right)\n",
    "    v_right_2 = backproject(P2, arrow_end_right)\n",
    "\n",
    "    n_left, d_left = find_plane_two_vectors(face_3d, v_left_1, v_left_2)\n",
    "    n_right, d_right = find_plane_two_vectors(c2, v_right_1, v_right_2)\n",
    "\n",
    "    c_, d_line = intersection_line_of_planes(n_left, d_left, n_right, d_right)\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "    closest_point, index, mds = find_closest_point(pointsd, c_, d_line)\n",
    "\n",
    "    image_module.set_img(im_left)\n",
    "    print(index)\n",
    "    image_module.show_box(left_frame_masks[index][\"bbox\"], out_path=f'long_video_out/frame_{i:04}')\n",
    "\n",
    "    image_module.set_img(im_right)\n",
    "    image_module.show_box(right_frame_masks[indexs[index]][\"bbox\"])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "final",
   "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": "3.10.16"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}