Skip to content
Branch: master
Go to file
Code

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
Feb 17, 2017
Feb 17, 2017
Mar 3, 2017
Mar 3, 2017

README.md

AnimeFace 2009

The face detector for anime/manga. This is similar to lbpcascade_animeface, but it's more accurate and supports facial landmark detection. I developed this library in 2008~2009.

Original website: http://anime.udp.jp/ (in Japanese)

demo

Figure (c) namco

Installation

Requirements

  • Ruby
  • ImageMagick
  • gcc, make

Ubuntu

sudo apt-get install libmagickwand-dev
sudo gem install rmagick
./build.sh

Arch Linux

Install imagemagick-full and ruby-rmagick from AUR, and then run ./build.sh.

Run sample code (Ruby)

cd animeface-ruby
ruby sample.rb <input image>

View at ${input_image}_out.png

Extract landmarks for a folder of images (Ruby)

First install Ruby package progress_bar. Then run:

cd animeface-ruby
ruby proc_folder.rb <input image folder> <output landmark file>

Each of the lines in the <output landmark file> is a JSON string corresponding to a file in <input image folder>. View the output file for more details.

Alternatively, replace proc_folder.rb with proc_folder_parallel.rb leads to image processing in parallel. This requires Ruby package parallel and ruby-progressbar

Create new dataset with animeface-ruby

  1. Prepare images first.
  2. Extract face images with animeface-ruby/face_collector.rb
face_collector.rb --src <image dir> --dest <output dir> --threshold <0.0~1.0, default: 0.2> --margin <0.0~, default: 0.1>
  1. Delete false positive images using windows explorer or something.
  2. Make annotation data from the filename (filename is formatted as ${orignal_file_name_without_extension}_${x}_${y}_${width}_${height}.png, see example)

About

Face and Landmark Detector for Anime/Manga. This is 2009s version of Imager::AnimeFace, but it works on recent system.

Topics

Resources

License

Releases

No releases published
You can’t perform that action at this time.