Type: Package
Title: Convolutional Neural Network for Face Detection
Description: An open source library for face detection in images. Provides a pretrained convolutional neural network based on https://github.com/ShiqiYu/libfacedetection which can be used to detect faces which have size greater than 10x10 pixels.
Maintainer: Jan Wijffels <jwijffels@bnosac.be>
License: BSD_3_clause + file LICENSE
Version: 0.1
URL: https://github.com/bnosac/image
Imports: Rcpp (≥ 0.12.8), graphics
LinkingTo: Rcpp
Suggests: magick
RoxygenNote: 7.1.0
NeedsCompilation: yes
Packaged: 2020-07-22 08:05:55 UTC; Jan
Author: Jan Wijffels [aut, cre, cph] (R wrapper), BNOSAC [cph] (R wrapper), Shiqi Yu [ctb, cph] (libfacedetection C++ code)
Repository: CRAN
Date/Publication: 2020-07-27 12:30:03 UTC

Detect faces in images using the libfacedetection CNN

Description

Detect faces in images using using a convolutional neural network available from https://github.com/ShiqiYu/libfacedetection. The function can be used to detect faces of minimal size 10x10 pixels.

Usage

image_detect_faces(x)

Arguments

x

an object of class magick-image with rgb colors. Or an rgb integer array with pixel values in the 0-255 range.

Value

A list with elements nr and detections.
Element nr indicates the number of faces found.
The data frame detections indicates the locations of these. This data.frame has columns x, y, width and height as well as a columns called confidence. The values of x and y are the top left of the start of the box. This data frame also has the x and y locations of 5 face landmarks (eyes, nose and mouth ends).

Examples

library(magick)
path <- system.file(package="image.libfacedetection", "images", "handshake.jpg")
x <- image_read(path)
x
faces <- image_detect_faces(x)
faces
plot(faces, x, border = "red", lwd = 7, col = "white", landmarks = TRUE)


##
## You can also directly pass on the RGB array in BGR format 
## without the need of having magick
##
tensor <- image_data(x, channels = "rgb")
tensor <- as.integer(tensor)
faces  <- image_detect_faces(tensor)
str(faces)
plot(faces, x)

Plot detected faces

Description

Plot functionality for bounding boxes detected with image_detect_faces

Usage

## S3 method for class 'libfacedetection'
plot(
  x,
  image,
  border = "red",
  lwd = 5,
  only_box = FALSE,
  col = "red",
  cex = 2,
  landmarks = FALSE,
  col_landmarks = "black",
  cex_landmarks = 1,
  pch_landmarks = 20,
  ...
)

Arguments

x

object of class libfacedetection as returned by image_detect_faces

image

object of class magick-image which was used to construct x

border

color of the border of the box. Defaults to red. Passed on to rect

lwd

line width of the border of the box. Defaults to 5. Passed on to rect

only_box

logical indicating to draw only the box and not the text on top of it. Defaults to FALSE.

col

color of the text on the box. Defaults to red. Passed on to text

cex

character expension factor of the text on the box. Defaults to 2. Passed on to text

landmarks

logical indicating to plot the landmarks as points. Defaults to FALSE.

col_landmarks

color of the point of the landmarks. Defaults to black.

cex_landmarks

cex of the point of the landmarks. Defaults to 1.

pch_landmarks

pch of the point of the landmarks. Defaults to 20.

...

other parameters passed on to rect

Value

an object of class magick-image

Examples

library(magick)
path <- system.file(package="image.libfacedetection", "images", "handshake.jpg")
x <- image_read(path)
x
faces <- image_detect_faces(x)
faces
plot(faces, x, border = "red", lwd = 7, col = "white")
plot(faces, x, border = "red", lwd = 7, col = "white", landmarks = TRUE, 
     col_landmarks = "purple", cex_landmarks = 2, pch_landmarks = 4)

## show one detected face
face <- head(faces$detections, 1)
image_crop(x, geometry_area(x = face$x, y = face$y, 
                            width = face$width, height = face$height))
## show all detected faces
boxcontent <- lapply(seq_len(faces$nr), FUN=function(i){
  face <- faces$detections[i, ]
  image_crop(x, geometry_area(x = face$x, y = face$y, 
                              width = face$width, height = face$height))
})
boxcontent <- do.call(c, boxcontent)
boxcontent