A free, privacy aware web application to redact faces from photos

The tool DFACE uses a facial recognition algorithm to find faces in any given photo and then blurs, redacts, or hides them with emojis according to the user's preference. Once the web application has loaded, it can work offline, meaning that photos are not stored, transmitted, or shared in any form. This is done using YOLOV5, a lightweight facial recognition architecture that detects any face bigger than 10x10 pixels and can be run directly in the user's browser, making the app accessible from any device.

Adam Harvey, the artist responsible for the concept and computer vision system, and Jules LaPlace, the software engineer behind the app, developed this project out of frustration over existing "face anonymization" tools. According to the two, other services neglect users' privacy, are costly, or do not work well. DFACE was developed as part of VFRAME, a project aimed at providing accessible computer vision tools to journalists, activists, and human rights researchers.

The reasons for redacting faces can be diverse, from complying with privacy regulations to protecting vulnerable individuals such as children or protesters. The creators themselves describe the tool as a measure of protection against the mass-surveillance users and citizens are often subject to online.

Similar projects to DFACE

Back to all projects.