Real time Virtual Background Desktop App (Windows only for now).
This app can read frames from real physical webcam/camera, replace it with virtual background. It could be used as a virtual camera source for other apps like Skype, Zoom, etc. Specially, it can run on CPU in real time. We got ~24 FPS on Intel core i5 9500F.
To archieve the goal, we combined several components.
PyQt5- GUIPytorch- Deep Learning model for background removal componentOpenCV- To read webcam/camera source.pyvirtualcam: To create virtual camera. See details at their home pagepygrabber: To get webcam/camera names.OpenVino: Deep Learning model optimization framework.
Note: Windows 10 only for now
Pre-built binary You can download the prebuilt version and use it directly without any effort.
From source In case you want to build it manually from source, please go through the following steps.
- Clone this repo to local:
git clone https://github.com/tamnguyenvan/beecam - Install OpenVino for Windows 10. This project only requires Python 3 (no need Visual Studio and CMake).
- Setup OpenVino environment variables.
- Install dependencies:
pip install -r requirements.txt. - Test it in Python:
python beecam.py
If everything is ok, let's build it using pyinstaller with the following command.
pyinstaller --onedir beecam.py \
--name beecam \
--add-binary "[OPENVINO_PATH]\python\python3.8\openvino\inference_engine;openvino/inference_engine" \
--add-data "config/;config/" \
--add-data "background.jpg;." \
--add-data "model.bin;." \
--add-data "model.xml;." \
--add-binary "[OPENVINO_PATH]\deployment_tools\inference_engine\bin\intel64\Release\*;." \
--add-binary "[OPENVINO_PATH]\deployment_tools\ngraph\lib;." \
--add-data "OBS-VirtualCam;OBS-VirtualCam" \
--ico beecam.ico --windowed --clean -y
Replace [OPENVINO_PATH] with your own OpenVino installation folder. The distribution should be appeared in dist folder.
- Clean code
- Reduce distribution size.
- Improve the deep learning model.
- More elegent UI