Face recognition example notebook

face_recognition.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "1b5aeccc-fe82-4225-9783-db5c89a557ca",
   "metadata": {},
   "source": [
    "# Required Installations\n",
    "\n",
    "## Installing dlib using `conda` with CUDA enabled\n",
    "\n",
    "The following instructions are copied from [https://gist.github.com/nguyenhoan1988/ed92d58054b985a1b45a521fcf8fa781](https://gist.github.com/nguyenhoan1988/ed92d58054b985a1b45a521fcf8fa781).\n",
    "\n",
    "Prerequisite: `conda` and/or `miniconda` are already installed. I would rather have `mamba` or `micromamba` for this.\n",
    "\n",
    "### Create a conda environment (optional, you may use your existing environment).\n",
    "\n",
    "```console\n",
    "$ conda create -n dlib python=3.9 cmake ipython\n",
    "```\n",
    "\n",
    "### Activate the environment (optional, you may use your existing environment).\n",
    "\n",
    "```console\n",
    "$ conda activate dlib\n",
    "```\n",
    "\n",
    "### Install CUDA and cuDNN with `conda` using nvidia channel.\n",
    "\n",
    "```console\n",
    "$ conda install cuda cudnn -c nvidia\n",
    "```\n",
    "\n",
    "Then find the path to the `nvcc` of this environment. We will use this path for the build step below, use the path without the suffix `nvcc`.\n",
    "\n",
    "```console\n",
    "$ which nvcc\n",
    "/path/to/your/miniconda3/envs/dlib/bin/nvcc\n",
    "```\n",
    "\n",
    "### Install dlib.\n",
    "\n",
    "Clone and build dlib from source. If you see a warning about CUBLAS use command `sudo apt-get install libopenblas-dev liblapack-dev` (on ubuntu systems).\n",
    "\n",
    "```console\n",
    "$ git clone https://github.com/davisking/dlib.git\n",
    "$ cd dlib\n",
    "$ mkdir build\n",
    "$ cd build\n",
    "$ cmake .. -DDLIB_USE_CUDA=1 -DUSE_AVX_INSTRUCTIONS=1 -DCUDAToolkit_ROOT=/path/to/your/miniconda3/envs/dlib/bin/\n",
    "$ cmake --build .\n",
    "$ cd ..\n",
    "$ python setup.py install --set DLIB_USE_CUDA=1\n",
    "```\n",
    "\n",
    "### Test dlib\n",
    "\n",
    "Test whether your installation is successful in the `ipython` terminal.\n",
    "\n",
    "```console\n",
    "(dlib) $ ipython\n",
    "Python 3.9.13 | packaged by conda-forge | (main, May 27 2022, 16:56:21) \n",
    "[GCC 10.3.0] on linux\n",
    "Type \"help\", \"copyright\", \"credits\" or \"license\" for more information.\n",
    "\n",
    "In [1]: import dlib\n",
    "\n",
    "In [2]: dlib.DLIB_USE_CUDA\n",
    "\n",
    "Out[2]: True\n",
    "\n",
    "In [3]: print(dlib.cuda.get_num_devices())\n",
    "\n",
    "1\n",
    "```\n",
    "\n",
    "## Install face_recognition library\n",
    "\n",
    "Run the following command or refer to the original image respository [https://github.com/ageitgey/face_recognition](https://github.com/ageitgey/face_recognition).\n",
    "```console\n",
    "$ pip install face_recognition\n",
    "```\n",
    "\n",
    "## Common error due to mismatching cuda versions\n",
    "\n",
    "You might observe the following in error when you have a conda installation and a system wide installation for cuda with different versions. For example `CUDA Version: 11.6` in `nvidia-smi` and `Cuda compilation tools, release 11.7, V11.7.99` in `nvcc --version`. In Ubuntu search for Additional drivers app and select the latest version `nvidia-driver-515` in this particular case. Use GUI in windows to update drivers (GEFORCE Experience). \n",
    "\n",
    "```\n",
    "cnn_face_detector = dlib.cnn_face_detection_model_v1('dlib_models/mmod_human_face_detector.dat')\n",
    "RuntimeError: Error while calling cudaMallocHost(&data, new_size*sizeof(float)) in ./dlib/dlib/cuda/gpu_data.cpp:211. code: 222, reason: the provided PTX was compiled with an unsupported toolchain.\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "b65ffeab-d6b2-4601-9250-e793906646a8",
   "metadata": {},
   "outputs": [],
   "source": [
    "import cv2\n",
    "import numpy as np\n",
    "import face_recognition\n",
    "import time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "cb790e1c-2ebe-4c24-b41f-b42077b9b634",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load a sample picture and learn how to recognize it.\n",
    "person_1 = face_recognition.load_image_file(\"./person-1.jpg\", mode='RGB')\n",
    "person_1_face_encoding = face_recognition.face_encodings(person_1)[0]\n",
    "\n",
    "# Load a second sample picture and learn how to recognize it.\n",
    "person_2_image = face_recognition.load_image_file(\"person-2.jpg\", mode='RGB')\n",
    "person_2_face_encoding = face_recognition.face_encodings(person_2_image)[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "3cb3f62f-f03a-454b-be8e-d9bca5402e4c",
   "metadata": {},
   "outputs": [],
   "source": [
    "known_face_encodings = [\n",
    "    person_1_face_encoding,\n",
    "    person_2_face_encoding\n",
    "]\n",
    "known_face_names = [\n",
    "    \"Name 1\",\n",
    "    \"Name 2\"\n",
    "]\n",
    "\n",
    "# Initialize some variables\n",
    "face_locations = []\n",
    "face_encodings = []\n",
    "face_names = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "1c5b9475-3d6c-4c34-bcdb-3289de637580",
   "metadata": {},
   "outputs": [],
   "source": [
    "video_capture = cv2.VideoCapture(0)\n",
    "# width = 1920\n",
    "# height = 1080\n",
    "# video_capture.set(cv2.CAP_PROP_FRAME_WIDTH, width)\n",
    "# video_capture.set(cv2.CAP_PROP_FRAME_HEIGHT, height)\n",
    "# video_capture.set(cv2.CAP_PROP_FPS, 60)\n",
    "scale = 2\n",
    "\n",
    "while True:\n",
    "    t1 = time.time()\n",
    "    # Grab a single frame of video\n",
    "    ret, frame = video_capture.read()\n",
    "    \n",
    "    if frame is None:\n",
    "        continue\n",
    "    \n",
    "    # Resize frame of video to 1/4 size for faster face recognition processing\n",
    "    small_frame = cv2.resize(frame, (0, 0), fx=1/scale, fy=1/scale)\n",
    "    \n",
    "    # Convert the image from BGR color (which OpenCV uses) to RGB color (which face_recognition uses)\n",
    "    rgb_small_frame = cv2.cvtColor(small_frame, cv2.COLOR_BGR2RGB)\n",
    "\n",
    "    # Find all the faces and face encodings in the current frame of video\n",
    "    face_locations = face_recognition.face_locations(rgb_small_frame)\n",
    "    face_encodings = face_recognition.face_encodings(rgb_small_frame, face_locations)\n",
    "\n",
    "    face_names = []\n",
    "    \n",
    "    for face_encoding in face_encodings:\n",
    "        # See if the face is a match for the known face(s)\n",
    "        matches = face_recognition.compare_faces(known_face_encodings, face_encoding)\n",
    "        name = \"Unknown\"\n",
    "\n",
    "        face_distances = face_recognition.face_distance(known_face_encodings, face_encoding)\n",
    "        best_match_index = np.argmin(face_distances)\n",
    "        if matches[best_match_index]:\n",
    "            name = known_face_names[best_match_index]\n",
    "\n",
    "        face_names.append(name)\n",
    "        \n",
    "    # Display the results\n",
    "    for (top, right, bottom, left), name in zip(face_locations, face_names):\n",
    "        # Scale back up face locations since the frame we detected in was scaled to 1/4 size\n",
    "        top *= scale\n",
    "        right *= scale\n",
    "        bottom *= scale\n",
    "        left *= scale\n",
    "\n",
    "        # Draw a box around the face\n",
    "        cv2.rectangle(frame, (left, top), (right, bottom), (0, 0, 255), 2)\n",
    "\n",
    "        # Draw a label with a name below the face\n",
    "        cv2.rectangle(frame, (left, bottom - 35), (right, bottom), (0, 0, 255), cv2.FILLED)\n",
    "        font = cv2.FONT_HERSHEY_DUPLEX\n",
    "        cv2.putText(frame, name, (left + 6, bottom - 6), font, 1.0, (255, 255, 255), 1)\n",
    "    \n",
    "    t2 = time.time()\n",
    "    fps = 1/(0.00001 + t2 - t1)\n",
    "    cv2.putText(frame, f\"{fps:3.1f} FPS\", (30, 30), cv2.FONT_HERSHEY_DUPLEX, 1.0, (255, 0, 0), 1)\n",
    "    # Display the resulting image\n",
    "    cv2.imshow('Video', frame)\n",
    "\n",
    "    # Hit 'q' on the keyboard to quit!\n",
    "    if cv2.waitKey(1) & 0xFF == ord('q'):\n",
    "        break\n",
    "        \n",
    "# Release handle to the webcam\n",
    "video_capture.release()\n",
    "cv2.destroyAllWindows()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}