Magritte is an image redaction library to disguise the identity of persons in photos and video streams.
Magritte offers Visual Deidentification technology that is
- robust: Magritte allows to use redaction technology that provides a higher level of obfuscation than simple blurring.
- reliable: We allow to combine and tweak detection strategies to detect sensitive content more reliably.
- fast: We offer solutions that are optimized towards a real-time situation, while others are optimized towards offline treatment.
- modular: You can combine various detection and redaction pieces in many different ways to find a solution that fits your situation, and you can also use them on their own.
WARNING: Magritte is currently in early development. We are still evaluating the precision of the detection and redaction methods we provide. For the face detection models, please refer to MediaPipe’s documentation to understand their precision.
If you want to try out some demos, see Trying out Magritte
If you’re an engineer and want to learn how to use Magritte in your project, we recommend starting with the codelab. It gives simple examples that help you understand the basic concepts and links to the more advanced material that you can read in case you need it.