face cage generator, 2017.
This is a Mac OSX application I assembled, which is designed as a tool for artists wishing to generate facial landmarks or "face cages" from images of people. The software allows one to take one or multiple images of faces and process them using contemporary facial recognition algorithms to produce the locations of "landmarks," or the pixel coordinates that define the features of a face (see images below). This is the same category of algorithm that companies like Apple use to face-unlock iPhones, or Amazon sells to police departments to implement their viciously biased predictive policing programs. This software does not require an internet connection to run and accordingly does not pass any information online in order to process face images. No company will gain access to your images by using this software.
Computer vision and facial recognition softwares are based on biometric technologies functionally similar to the phrenology and race-science of former eugenic regimes. Despite the misconception that they are purely objective, the underlying logic of facial recognition posits identity as legible on the surface of an individual (read Simone Browne's concept of Digital Epidermalization), such that they frequently demarcate marginalized peoples (e.g. queer people, people of color) as abnormal, often subjecting such individuals to increased scrutiny, harassment, or violence depending on the implementation. This type of technology is increasingly finding use in both consumer and military software, helping surveil those who are deemed to be especially "high risk" It can at times be necessary to examine or utilize the tools of state surveillance in order to understand, subvert, or otherwise disrupt them. This software is designed to make the use of such technology more accessible to artists that may wish to utilize it in their projects.
This application was initially created for Gerri Ondrizek's project on the use of biometric technologies to surveil refugee groups fleeing Syria, but is free for use by anybody. The facial landmarks are generated using the dlib facial landmark detection algorithm, and the face cages are created using an algorithm known as Delaunay Triangulation. Thanks to the learnopencv.com site created by Dr. Satya Mallick for providing much of the code base for this application.
Click here to download the application.