Git Product home page Git Product logo

obs-realsense-greenscreen's Introduction

OBS RealSense Virtual Greenscreen

This OBS plugin allows to use the depth sensor of the Intel RealSense USB cameras to generate a virtual greenscreen. The camera can already be used OBS directly as a normal camera (via V4L2) but this does not use the depth sensor.

With this plugin the camera can be used directly in OBS without OS servoces. The actual access to the camera and the alignment of the video with the depth field is done by the librealsense2 library. This plugin just performs the masking and it implements the interface to OBS.

Building

If the required packages are installed, just run make. The include RPM .spec file codifies the dependencies which are

  • gcc 10 or higher (tested with C++20 only)
  • librealsense. Fedora 33 ships the old 2.38 version but it worked for me. Fedora 35 ships 2.40 which also works.
  • gtkmm 3.

For both librealsense and gtkmm it is of course necessary to install the -devel sub-package as well.

Alternative librealsense

To test a librealsense other than the officially installed version use the make command with the variable ALT_LIBREALSENSE pointing to the appropriate directory:

make ALT_LIBREALSENSE=/home/drepper/librealsense

This works with the official librealsense sources off of github.

Testing

Before running OBS with the plugin, when all other things can potentially go wrong, use the testrealsense binary. It consists of just a few lines of C++ using gtkmm to create an appropriately sized window and display the frames the camera provides. Nothing else. Press the escape key to exit the program. The progam is not build and shipped when you use the include RPM .spec file.

Using the plugin with OBS

After installation the plugin should be immediately found by OBS. Adding it as a source happens just as for other cameras, just select the "RealSense Greenscreen" source.

The property dialog allows to select the device, change the resolution, set the maximum distance (in meters), the size of the depth filter, and greenscreen color.

After the camera source has been added one can use the chroma key filter. To enable the filter select the RealSense Greenscreen source in the Sources list. Right click on the entry to bring up the context dialog and select the Filters menu item. This allows to add the Chroma Key effect filter. Just make sure to select the key color as selected for the source and play a bit with the Similarity and Key Color Spill Reduction sliders to get an acceptable result.

Filtering the depth field happens by taking the average value for the last N values. N=1 means no filtering. Any larger value will increase the CPU time and memory needed but it will reduce the effects of the noisy depth field sensor. This sensor frequently produces at random location a non-sensical value and simple averiging seems to be sufficient to take care of this. The default value is four, meaning the average value of the previous four frames is used.

Caveats

This plugin has so far been tested only on my machine:

  • The only camera tested so far is the L515. I hope that the librealsense2 library handles the other devices the same.
  • I have not even tested what happens if no camera is attached.
  • The frequency of the picture provided is fixed at 30Hz. Not sure whether this needs to be variable.
  • The depth sensor is quite noisy. Select an appropriate depth filter size. The larger the size, the more work is required, so restrict the size to a reasonable value for your machine.
  • The code allows to set a minimum distance for the mask as well. This setting is so far not carried through to the OBS properties dialog. Not sure this is necessary.
  • Obviously, the code has only been tested under Fedora.

Trouble Shooting

Connecting to the camera sometimes fails. One will see then errors from librealsense like

RealSense error calling rs2_pipeline_wait_for_frames(pipe:0x1ffabe0):
    Frame didn't arrive within 15000

For me this does not mean a problem in the code. Instead, it seems related to the USB device connection. Unplugging the camera and then reattaching it usually fixes this for me.

In future I might try to automatically handle this. Linux allows to reset an USB device through an ioctl call with USBDEVFS_RESET. I will give this a try the next time I run into the problem.

If the camera is already in use the plugin will not report any resources to OBS. Any already source in a scene will just be zero-sized.

Author

Ulrich Drepper [email protected]

Parts of the realsense-greenscreen.cc are derived from the rs-align-advanced example in Intel's librealsense which has the copyright notice

 // Copyright(c) 2017 Intel Corporation. All Rights Reserved.
 // License: Apache 2.0. See LICENSE file in root directory.

obs-realsense-greenscreen's People

Contributors

drepper avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar

Forkers

4aiman

obs-realsense-greenscreen's Issues

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.