Face Detection in Elixir

detect.livemd

Face Detection in Elixir

Mix.install(
  [
    {:evision, "~> 0.1"},
    {:kino, "~> 0.7"}
  ],
  system_env: [
    # optional, defaults to `true`
    # set `EVISION_PREFER_PRECOMPILED` to `false`
    # if you prefer `:evision` to be compiled from source
    # note that to compile from source, you may need at least 1GB RAM
    {"EVISION_PREFER_PRECOMPILED", true},

    # optional, defaults to `true`
    # set `EVISION_ENABLE_CONTRIB` to `false`
    # if you don't need modules from `opencv_contrib`
    {"EVISION_ENABLE_CONTRIB", true},

    # optional, defaults to `false`
    # set `EVISION_ENABLE_CUDA` to `true`
    # if you wish to use CUDA related functions
    # note that `EVISION_ENABLE_CONTRIB` also has to be `true`
    # because cuda related modules come from the `opencv_contrib` repo
    {"EVISION_ENABLE_CUDA", false},

    # required when
    # - `EVISION_ENABLE_CUDA` is `true`
    # - and `EVISION_PREFER_PRECOMPILED` is `true`
    #
    # set `EVISION_CUDA_VERSION` to the version that matches
    # your local CUDA runtime version
    #
    # current available versions are
    # - 118
    # - 121
    {"EVISION_CUDA_VERSION", "118"},

    # require for Windows users when
    # - `EVISION_ENABLE_CUDA` is `true`
    # set `EVISION_CUDA_RUNTIME_DIR` to the directory that contains
    # CUDA runtime libraries
    {"EVISION_CUDA_RUNTIME_DIR", "C:/PATH/TO/CUDA/RUNTIME"}
  ]
)

Camera Initialization

capture = Evision.VideoCapture.videoCapture(0)

face_cascade_path =
  Path.join([
    :code.priv_dir(:evision),
    "share/opencv4/haarcascades/haarcascade_frontalface_default.xml"
  ])

face_cascade = Evision.CascadeClassifier.cascadeClassifier(face_cascade_path)

Do it Once

frame = Evision.VideoCapture.read(capture) |> Evision.resize({1280, 720})

grey = Evision.cvtColor(frame, Evision.Constant.cv_COLOR_BGR2GRAY())

faces =
  Evision.CascadeClassifier.detectMultiScale(
    face_cascade,
    grey,
    scaleFactor: 1.8,
    minNeighbors: 4
  )

# Draw a red rectangle over each detected face
mat =
  Enum.reduce(faces, frame, fn {x, y, w, h}, mat ->
    Evision.rectangle(mat, {x, y}, {x + w, y + h}, {0, 0, 255}, thickness: 2)
  end)

Livebook Frames

# Establish a frame
# You can have as many of these as you want
# They could even contain form elements
out = Kino.Frame.new()
Kino.render(out)

Enum.each(0..10, fn x ->
  # Update the frame while the code runs
  # This could be something like adding data to a chart, for example
  Kino.Frame.append(out, x)
  :timer.sleep(100)
end)

Kino.Frame.render(out, "Done")

Do it live

out = Kino.Frame.new()
Kino.render(out)

for _ <- 0..100 do
  frame =
    Evision.VideoCapture.read(capture)
    |> Evision.resize({640, 360})

  grey_frame = Evision.cvtColor(frame, Evision.Constant.cv_COLOR_BGR2GRAY())

  faces =
    Evision.CascadeClassifier.detectMultiScale(
      face_cascade,
      grey_frame,
      scaleFactor: 1.8,
      minNeighbors: 4
    )

  # Draw a red rectangle over each detected face
  mat =
    Enum.reduce(faces, frame, fn {x, y, w, h}, mat ->
      Evision.rectangle(mat, {x, y}, {x + w, y + h}, {0, 0, 255}, thickness: 2)
    end)

  target_x = 300
  target_y = 300

  # Find the center of all detected faces
  {center_x, center_y} =
    if Enum.empty?(faces) do
      {2, [h, w]} = Evision.Mat.size(frame)
      {w / 2, h / 2}
    else
      x = Enum.sum(Enum.map(faces, fn {x, _y, w, _h} -> x + w / 2 end)) / length(faces)
      y = Enum.sum(Enum.map(faces, fn {_x, y, _w, h} -> y + h / 2 end)) / length(faces)

      {x, y}
    end

  # Find a 300x300 rectangle to crop to
  # Ideally, centered around the face
  bounding_box = {
    max(0, round(center_x - target_x / 2)),
    max(0, round(center_y - target_y / 2)),
    target_x,
    target_y
  }

  # Crop the image down to the 300x300 box
  mat = Evision.Mat.roi(mat, bounding_box)

  Kino.Frame.render(out, mat)
end

:ok

Cool