Description

Benjamin Elder published the great article Open Source Virtual Background which explained how to create a camera feed with a virtual background for use in conferencing applications. While the article itself is published here on GitHub, unfortunately, the code itself wasn't gathered together in a repository to make it easy for others to use. Additionally, the article was written for a Debian derivative. My machines run Fedora and CentOS so I put together this repository to make it easier for people running Fedora and CentOS to be able to replicate Ben's work.

Notes

• TCP port 9000 is used to connect the container running bodypix to the container running the fake camera over a bridged network which is automatically created.
• The camera output is limited (to improve performance) to 1280x720.
• The containers used are named bodypix and fakecam.

Requirements

• Fedora 31 or CentOS 7.7
• Docker CE 19.03.8
• Optional: NVIDIA graphics card with CUDA support, for faster inference
  • NVIDIA GPU drivers 418.x or later
  • CUDA toolkit 10.1

Install

1. Clone the repository into fakecam directory:

   git clone ... fakecam
   cd fakecam

2. Run setup:

   ./setup

3. Copy the image you want to use to fakecam/background.jpg. For example:

   cp backgrounds/star-wars-feature-vf-2019-summer-embed-07.jpg \
     fakecam/background.jpg

4. Build the Docker images:

   ./build

Use

Run vbackcam to start and stop the fake camera with the virtual background.

While the fake camera is running, open your conferencing application and select v4l2loopback or /dev/video20 as the camera.

Change the background

1. Copy the new background image into fakecam/background.jpg
2. Rebuild the fakecam Docker image:

   docker build --tag fakecam ./fakecam
   

Useful links

• Add kernel modules with DKMS
• Backgrounds:
  • Star Wars: The Rise of Skywalker Photos
  • TrekCore.com Gallery
  • Unsplash: The internet's source of freely usable images
• BodyPix: Real-time Person Segmentation in the Brower with TensorFlow.js
  • BodyPix repository
  • body-pix npm
• Docker Engine
  • Install Docker Engine on CentOS
  • Install Docker Engine on Fedora
• node.js
• NVIDIA
  • CUDA enabled GPUs
  • CUDA toolkit
  • cuDNN: CUDA Deep Neural Network library
  • GPU drivers
  • nvidia-docker: Build and run Docker containers leveraging NVIDIA GPUs
• Open Source Virtual Background
• OpenCV python bindings
• pyfakewebcam: A library for writing RGB frames to a fake webcam device on Linux
• TensorFlow: Open source library to help deliver and train ML models
  • TensorFlow GPU support
  • TensorFlow.js: A library for machine learning in JavaScript
• Video for Linux Two (V4L2)
  • Linux TV
    • Part I - Video for Linux API
  • v4l2loopback: A kernel module to create V4L2 loopback devices
  • v4l2ucp: Universal control panel for Video for Linux Two (V4L2) devices

License

Photo copyrights:

• Annie Leibovitz or Vanity Fair/Condé Nast
• Dorsa Masghati on Unsplash wallpapers
• Moosa Haleem on Unsplash wallpapers
• Nastuh Abootalebi on Unsplash interiors
• TrekCore.com

Copyright © 2020 Christopher Voltz

This package is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

This package is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.