usingMediaPipeTFlite.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Internet Example of IRIS alone"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "input_details [{'name': 'input_1', 'index': 0, 'shape': array([ 1, 64, 64,  3]), 'shape_signature': array([ 1, 64, 64,  3]), 'dtype': <class 'numpy.float32'>, 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array([], dtype=float32), 'zero_points': array([], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}]\n",
      "output_details [{'name': 'output_eyes_contours_and_brows', 'index': 384, 'shape': array([  1, 213]), 'shape_signature': array([  1, 213]), 'dtype': <class 'numpy.float32'>, 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array([], dtype=float32), 'zero_points': array([], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}, {'name': 'output_iris', 'index': 385, 'shape': array([ 1, 15]), 'shape_signature': array([ 1, 15]), 'dtype': <class 'numpy.float32'>, 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array([], dtype=float32), 'zero_points': array([], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}]\n",
      "is floating model: True\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "\n",
    "img = np.random.randint(0, 256, (1,64,64,3), np.uint8)\n",
    "\n",
    "img = (np.float32(img) - 0.0) / 255.0 # normalization (specified in tflite_converter_calculator, not in model card)\n",
    "\n",
    "model_file = 'iris_landmark.tflite'\n",
    "interpreter = tf.lite.Interpreter(\n",
    "    model_path=model_file)\n",
    "interpreter.allocate_tensors()\n",
    "\n",
    "input_details = interpreter.get_input_details()\n",
    "output_details = interpreter.get_output_details()\n",
    "\n",
    "print(f'input_details {input_details}')\n",
    "print(f'output_details {output_details}')\n",
    "\n",
    "# check the type of the input tensor\n",
    "floating_model = input_details[0]['dtype'] == np.float32\n",
    "print(f'is floating model: {floating_model}')\n",
    "\n",
    "# NxHxWxC, H:1, W:2\n",
    "height = input_details[0]['shape'][1]\n",
    "width = input_details[0]['shape'][2]\n",
    "\n",
    "interpreter.set_tensor(input_details[0]['index'], img)\n",
    "\n",
    "interpreter.invoke()\n",
    "\n",
    "output_eyes = interpreter.get_tensor(output_details[0]['index'])[0]\n",
    "output_iris = interpreter.get_tensor(output_details[1]['index'])[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "output iris landmarks x [0.6636077  0.76640826 0.6471603  0.5568398  0.67772526]\n",
      "output iris landmarks y [0.51719385 0.50089335 0.41648433 0.53332144 0.6193517 ]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1dda6e13f10>]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import cv2\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "image = cv2.imread('000078ZOOMED.jpg')\n",
    "plt.imshow(image[:,:,::-1])\n",
    "\n",
    "image_cropped = image[:image.shape[0], :image.shape[0], ::-1] # crop to avoid letterboxing step\n",
    "plt.imshow(image_cropped)\n",
    "\n",
    "img = cv2.resize(image_cropped, (64, 64))[np.newaxis, :, :, :]\n",
    "img = (np.float32(img) - 0.0) / 255.0  # normalization (specified in tflite_converter_calculator, not in model card)\n",
    "\n",
    "#img = (np.float32(img) - 0.0) / 255.0 # normalization (specified in tflite_converter_calculator, not in model card)\n",
    "interpreter.set_tensor(input_details[0]['index'], img)\n",
    "\n",
    "interpreter.invoke()\n",
    "\n",
    "output_eyes = interpreter.get_tensor(output_details[0]['index'])[0]\n",
    "output_iris = interpreter.get_tensor(output_details[1]['index'])[0]\n",
    "\n",
    "print(f'output iris landmarks x {output_iris[::3]/64.0}') # tflite_tensors_to_landmarks_calculator normalization\n",
    "print(f'output iris landmarks y {output_iris[1::3]/64.0}')\n",
    "\n",
    "# cropped image\n",
    "plt.imshow(image_cropped)\n",
    "plt.plot(output_iris[::3]/64.0*image_cropped.shape[0], (1-output_iris[1::3]/64.0)*image_cropped.shape[1], '*')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1dda6eb40d0>]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# original image\n",
    "plt.imshow(image[:,:,::-1])\n",
    "plt.plot(output_iris[::3]/64.0*image_cropped.shape[0], (1-output_iris[1::3]/64.0)*image_cropped.shape[1], '*')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Using Library\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fdlite import FaceDetection, FaceDetectionModel\n",
    "from fdlite.render import Colors, detections_to_render_data, render_to_image \n",
    "from PIL import Image\n",
    "\n",
    "image = Image.open('books/smallme.jpg')\n",
    "detect_faces = FaceDetection(model_type=FaceDetectionModel.FRONT_CAMERA)\n",
    "faces = detect_faces(image)\n",
    "if not len(faces):\n",
    "    print('no faces detected :(')\n",
    "else:\n",
    "    render_data = detections_to_render_data(faces, bounds_color=Colors.GREEN)\n",
    "    render_to_image(render_data, image).show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "pass an image name as argument\n"
     ]
    }
   ],
   "source": [
    "from bisect import bisect_left, bisect_right\n",
    "from os import PathLike\n",
    "import numpy as np\n",
    "from PIL import Image, ImageOps\n",
    "from PIL.Image import Image as PILImage\n",
    "import sys\n",
    "from typing import Sequence, Tuple, Union\n",
    "from fdlite.face_detection import FaceDetection, FaceDetectionModel\n",
    "from fdlite.face_landmark import FaceLandmark, face_detection_to_roi\n",
    "from fdlite.iris_landmark import IrisLandmark, IrisResults\n",
    "from fdlite.iris_landmark import iris_roi_from_face_landmarks\n",
    "from fdlite.transform import bbox_from_landmarks\n",
    "\"\"\"Iris recoloring example based on face- and iris detection models.\n",
    "\"\"\"\n",
    "\n",
    "_Point = Tuple[int, int]\n",
    "_Size = Tuple[int, int]\n",
    "_Rect = Tuple[int, int, int, int]\n",
    "\n",
    "\n",
    "def recolor_iris(\n",
    "    image: PILImage,\n",
    "    iris_results: IrisResults,\n",
    "    iris_color: Tuple[int, int, int]\n",
    ") -> PILImage:\n",
    "    \"\"\"Colorize an eye.\n",
    "    Args:\n",
    "        image (Image): PIL image instance containing a face.\n",
    "        iris_results (IrisResults): Iris detection results.\n",
    "        iris_color (tuple): Tuple of `(red, green, blue)` representing the\n",
    "            new iris color to apply. Color values must be integers in the\n",
    "            range [0, 255].\n",
    "    Returns:\n",
    "        (Image) The function returns the modified PIL image instance.\n",
    "    \"\"\"\n",
    "    iris_location, iris_size = _get_iris_location(iris_results, image.size)\n",
    "    # nothing fancy - just grab the iris part as an Image and work with that\n",
    "    eye_image = image.transform(iris_size, Image.EXTENT, data=iris_location)\n",
    "    eye_image = eye_image.convert(mode='L')\n",
    "    eye_image = ImageOps.colorize(eye_image, 'black', 'white', mid=iris_color)\n",
    "    # build a mask for copying back into the original image\n",
    "    # no fancy anti-aliasing or blending, though\n",
    "    mask = _get_iris_mask(iris_results, iris_location, iris_size, image.size)\n",
    "    image.paste(eye_image, iris_location, mask)\n",
    "    return image\n",
    "\n",
    "\n",
    "def _get_iris_location(\n",
    "    results: IrisResults, image_size: _Size\n",
    ") -> Tuple[_Rect, _Size]:\n",
    "    \"\"\"Return iris location and -size\"\"\"\n",
    "    bbox = bbox_from_landmarks(results.iris).absolute(image_size)\n",
    "    width, height = int(bbox.width + 1), int(bbox.height + 1)\n",
    "    size = (width, height)\n",
    "    left, top = int(bbox.xmin), int(bbox.ymin)\n",
    "    location = (left, top, left + width, top + height)\n",
    "    return location, size\n",
    "\n",
    "\n",
    "def _get_iris_mask(\n",
    "    results: IrisResults,\n",
    "    iris_location: _Rect,\n",
    "    iris_size: _Size,\n",
    "    image_size: _Size\n",
    ") -> PILImage:\n",
    "    \"\"\"Return a mask for the visible portion of the iris inside eye landmarks.\n",
    "    \"\"\"\n",
    "    left, top, _, bottom = iris_location\n",
    "    iris_width, iris_height = iris_size\n",
    "    img_width, img_height = image_size\n",
    "    # sort lexicographically by x then y\n",
    "    eyeball_sorted = sorted([(int(pt.x * img_width), int(pt.y * img_height))\n",
    "                             for pt in results.eyeball_contour])\n",
    "    bbox = bbox_from_landmarks(results.eyeball_contour).absolute(image_size)\n",
    "    x_ofs = left\n",
    "    y_ofs = top\n",
    "    y_start = int(max(bbox.ymin, top))\n",
    "    y_end = int(min(bbox.ymax, bottom))\n",
    "    mask = np.zeros((iris_height, iris_width), dtype=np.uint8)\n",
    "    # iris ellipse radii (horizontal and vertical radius)\n",
    "    a = iris_width // 2\n",
    "    b = iris_height // 2\n",
    "    # iris ellipse foci (horizontal and vertical)\n",
    "    cx = left + a\n",
    "    cy = top + b\n",
    "    box_center_y = int(bbox.ymin + bbox.ymax) // 2\n",
    "    b_sqr = b**2\n",
    "    for y in range(y_start, y_end):\n",
    "        # evaluate iris ellipse at y\n",
    "        x = int(a * np.math.sqrt(b_sqr - (y-cy)**2) / b)\n",
    "        x0, x1 = cx - x, cx + x\n",
    "        A, B = _find_contour_segment(eyeball_sorted, (x0, y))\n",
    "        left_inside = _is_below_segment(A, B, (x0, y), box_center_y)\n",
    "        C, D = _find_contour_segment(eyeball_sorted, (x1, y))\n",
    "        right_inside = _is_below_segment(C, D, (x1, y), box_center_y)\n",
    "        if not (left_inside or right_inside):\n",
    "            continue\n",
    "        elif not left_inside:\n",
    "            x0 = int(max((B[0] - A[0])/(B[1] - A[1]) * (y - A[1]) + A[0], x0))\n",
    "        elif not right_inside:\n",
    "            x1 = int(min((D[0] - C[0])/(D[1] - C[1]) * (y - C[1]) + C[0], x1))\n",
    "        # mark ellipse row as visible\n",
    "        mask[(y - y_ofs), int(x0 - x_ofs):int(x1 - x_ofs)] = 255\n",
    "    return Image.fromarray(mask, mode='L')\n",
    "\n",
    "\n",
    "def _is_below_segment(A: _Point, B: _Point, C: _Point, mid: int) -> bool:\n",
    "    \"\"\"Return whether a point is below the line AB\"\"\"\n",
    "    dx = B[0] - A[0]\n",
    "    dy = B[1] - A[1]\n",
    "    if not dx:\n",
    "        # vertical line: check if the point is on the line\n",
    "        return A[1] <= C[1] and B[1] >= C[1]\n",
    "    # x -> [0, 1]\n",
    "    x = (C[0] - A[0]) / dx\n",
    "    m = dy / dx\n",
    "    y = x * m + A[1]\n",
    "    # flip the sign if the leftmost point is below the threshold\n",
    "    sign = -1 if A[1] > mid else 1\n",
    "    return sign * (C[1] - y) > 0\n",
    "\n",
    "\n",
    "def _find_contour_segment(\n",
    "    contour: Sequence[_Point], point: _Point\n",
    ") -> Tuple[_Point, _Point]:\n",
    "    \"\"\"Find contour segment (points A and B) that contains the point.\n",
    "    (contour must be lexicographically sorted!)\n",
    "    \"\"\"\n",
    "    def distance(a: _Point, b: _Point) -> int:\n",
    "        return (a[0] - b[0]) ** 2 + (a[1] - b[1]) ** 2\n",
    "\n",
    "    MAX_IDX = len(contour)-1\n",
    "    # contour point left of point\n",
    "    left_idx = max(bisect_left(contour, point) - 1, 0)\n",
    "    right_idx = min(bisect_right(contour, point), MAX_IDX)\n",
    "    d = distance(point, contour[left_idx])\n",
    "    # try to find a closer left point\n",
    "    while left_idx > 0 and d > distance(point, contour[left_idx - 1]):\n",
    "        left_idx -= 1\n",
    "        d = distance(point, contour[left_idx - 1])\n",
    "    # try to find a closer right point\n",
    "    d = distance(point, contour[right_idx])\n",
    "    while right_idx < MAX_IDX and d > distance(point, contour[right_idx + 1]):\n",
    "        right_idx += 1\n",
    "        d = distance(point, contour[right_idx + 1])\n",
    "    return (contour[left_idx], contour[right_idx])\n",
    "\n",
    "\n",
    "def main(image_file: Union[str, PathLike]) -> None:\n",
    "    EXCITING_NEW_EYE_COLOR = (161, 52, 216)\n",
    "    img = Image.open(image_file)\n",
    "    # run the detection pipeline\n",
    "    # Step 1: detect the face and get a proper ROI for further processing\n",
    "    face_detection = FaceDetection(FaceDetectionModel.BACK_CAMERA)\n",
    "    detections = face_detection(img)\n",
    "    if not len(detections):\n",
    "        print('No face detected :(')\n",
    "        exit(0)\n",
    "    face_roi = face_detection_to_roi(detections[0], img.size)\n",
    "\n",
    "    # Step 2: detect face landmarks and extract per-eye ROIs from them\n",
    "    face_landmarks = FaceLandmark()\n",
    "    landmarks = face_landmarks(img, face_roi)\n",
    "    eyes_roi = iris_roi_from_face_landmarks(landmarks, img.size)\n",
    "\n",
    "    # Step 3: perform iris detection to get detailed landmarksfor each eye\n",
    "    iris_landmarks = IrisLandmark()\n",
    "    left_eye_roi, right_eye_roi = eyes_roi\n",
    "    left_eye_results = iris_landmarks(img, left_eye_roi)\n",
    "    right_eye_results = iris_landmarks(img, right_eye_roi, is_right_eye=True)\n",
    "\n",
    "    # Step 4: apply new eye and exciting eye color\n",
    "    recolor_iris(img, left_eye_results, iris_color=EXCITING_NEW_EYE_COLOR)\n",
    "    recolor_iris(img, right_eye_results, iris_color=EXCITING_NEW_EYE_COLOR)\n",
    "\n",
    "    # Step 5: Profit! ...or just show the result ¯\\_(ツ)_/¯\n",
    "    img.show()\n",
    "\n",
    "\n",
    "if __name__ == '__main__':\n",
    "    if len(sys.argv) != 2:\n",
    "        print('pass an image name as argument')\n",
    "    else:\n",
    "        main(sys.argv[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "EXCITING_NEW_EYE_COLOR = (161, 52, 216)\n",
    "img = Image.open(\"books/metest.jpg\")\n",
    "    # run the detection pipeline\n",
    "    # Step 1: detect the face and get a proper ROI for further processing\n",
    "face_detection = FaceDetection(FaceDetectionModel.BACK_CAMERA)\n",
    "detections = face_detection(img)\n",
    "if not len(detections):\n",
    "        print('No face detected :(')\n",
    "        exit(0)\n",
    "face_roi = face_detection_to_roi(detections[0], img.size)\n",
    "\n",
    "    # Step 2: detect face landmarks and extract per-eye ROIs from them\n",
    "face_landmarks = FaceLandmark()\n",
    "landmarks = face_landmarks(img, face_roi)\n",
    "eyes_roi = iris_roi_from_face_landmarks(landmarks, img.size)\n",
    "\n",
    "    # Step 3: perform iris detection to get detailed landmarksfor each eye\n",
    "iris_landmarks = IrisLandmark()\n",
    "left_eye_roi, right_eye_roi = eyes_roi\n",
    "left_eye_results = iris_landmarks(img, left_eye_roi)\n",
    "right_eye_results = iris_landmarks(img, right_eye_roi, is_right_eye=True)\n",
    "\n",
    "    # Step 4: apply new eye and exciting eye color\n",
    "recolor_iris(img, left_eye_results, iris_color=EXCITING_NEW_EYE_COLOR)\n",
    "recolor_iris(img, right_eye_results, iris_color=EXCITING_NEW_EYE_COLOR)\n",
    "\n",
    "    # Step 5: Profit! ...or just show the result ¯\\_(ツ)_/¯\n",
    "img.show()\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.9.12 ('base')",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.12"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "80dc181fd9c4c04679f39f30cc343a2aa164fc30fbb63db40714068b01465031"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}