floe/backscrub
A virtual video device that applies deep semantic segmentation to replace video backgrounds in real time.
Not currently ranked — collecting fresh signals.
star history
BackScrub creates a virtual webcam device that uses deep learning models to perform semantic segmentation on video frames, identifying people and replacing the background with an image. It leverages TensorFlow Lite, MediaPipe, and BodyPix models compiled to TFLite for efficient inference. The application runs as a userspace V4L2 camera source, outputting processed frames to any video conferencing software.
Frequently asked
- What is floe/backscrub?
- A virtual video device that applies deep semantic segmentation to replace video backgrounds in real time.
- Is backscrub open source?
- Yes — floe/backscrub is open source, released under the Apache-2.0 license.
- What language is backscrub written in?
- floe/backscrub is primarily written in C++.
- How popular is backscrub?
- floe/backscrub has 746 stars on GitHub.
- Where can I find backscrub?
- floe/backscrub is on GitHub at https://github.com/floe/backscrub.