Unmasking Facial Recognition, Ways of [Machine] Seeing - Shinji Toya
<Instructions for the mask-making session>
Form groups of 3-5 to work with one laptop for each group.
Set up the laptop with the face detection demo loaded on a browser and a webcam. The demo page runs detection of a face when the browser is given access to the webcam. To run a more robust detection, change the face detector to “SSD Mobilenet V1” if it is feasible.
In groups, create examples of wearable masks that obfuscate the face detection technology. Participants can take turns so that they can all participate in the activity. They will be asked to observe and take note of how the interference occurs and any effective approaches.
>> Nominate a mask model or a wearer for the prototype. The face detection demo will run on this person while prototyping.
>> Use the mask-making materials to collaboratively build a wearable until the face detection can no longer detect a face in the webcam feed.
>> The wearer may move around to test the robustness of the prototype, i.e. ensuring no face will be detected from different angles.
>> Document the process and outcomes of the prototype in screenshots and photographs (i.e. if the participants have a smartphone or a camera)
Webcam face detection - faceapi.js : Github repo shared with MIT Licence by 2018 Vincent Mühler
https://justadudewhohacks.github.io/face-api.js/webcam_face_tracking
Paint Your Face Away, Shinji Toya
Paint Your Face Away Version 1.3 (works best on computers)
https://paintyourfaceaway.net/_ver/1.31d/
Resources compiled by Shinji Toya