face-extractor

extract faces as png images from files

Installation

Install dependencies

OSX

brew install pkg-config cairo pango libpng jpeg giflib librsvg unoconv

Ubuntu

sudo apt-get install pkg-config libcairo2-dev libpango-1.0-0 libpng-dev libjpeg-dev libgif-dev librsvg2-bin unoconv

Install from npm registry

npm install -g face-extractor

Usage

face-extractor /path/to/file.docx --output /path/to/output/folder --tensorflow --background "rgba(0, 0, 0, 0)" --square --pad 100

this will extract all detected faces in all images in given document and save them to /path/to/output/folder/face-${imageIndex}-${faceIndex}.png

Using Tensorflow

If you want to use tensorflow native bindings you have to install them globally like so:

npm install -g @tensorflow/tfjs-node@1.7.4

Note that we have to specify version lower than 2 because face-api.js is not compatible with newer versions of tensorflow.

Options

• --input (Required): a string pointing to the desired input file
• --output (Required): a string pointing to the desired output folder
• --background: a string representing the desired background color (if extracted face needs to be filled with a background). i.e. #000, rgba(255, 100, 50, 0.5), etc.
• --square: if passed, extracts a square image from detected faces
• --pad: a number, defining how much spacing should be added around the detected face (value is defined as a percentage, based on the size of the detected face. i.e. a value of 100 would add double the width and height to the image than it is actually detected)
• --tensorflow: if passed, uses native bindings (speeds things up dramatically)

Supported File types

• doc
• docx
• odt
• pdf
• png
• jpeg
• jpg