Simon Willison reported on a new experimental utility called Photo Scrubber, which is designed to remove metadata and automatically blur identifiable faces in digital photographs. The tool was built using an advanced model to handle face detection locally on the device.
Technology Behind the Tool
The application relies on Google’s MediaPipe C++ library, which has been compiled to WebAssembly via the tasks-vision package. It specifically integrates the BlazeFace face detection model to locate individuals within an image. By processing photographs locally, the software avoids uploading sensitive visual data to external servers, addressing privacy concerns associated with sharing images of strangers online.
Unknown Technical Specifications
The announcement did not specify the exact performance benchmarks of the WebAssembly implementation across different hardware configurations. Furthermore, details regarding supported image file formats, the full scope of removed metadata types, and the exact release timeline or availability of the source code were not disclosed.
As privacy tools evolve, local execution models like this offer a promising approach to handling sensitive visual data securely in the browser.
Lead image: Alexas Fotos / Pexels
