A Python script to remove background of a webcam image and stream the result into a virtual video device. The background removal is based on face detection with OpenCV within each image.
- Tested with Python 3.7
- Requires V4L2-Loopback.
In Ubuntu this is available as the package
v4l2loopback-utilsand its dependencies. HOWEVER under Ubuntu 18.04 with recent Kernel, the module has a bug (see here also). Hence, it is advised to installv4l2loopbackfrom github source.
- create a virtual video device
$ sudo modprobe v4l2loopback devices=1 video_nr=20 card_label='v4l2loopback' exclusive_caps=1 - Make sure you the video device
/dev/video20has been created. - Run
python removeCamBg.py
See python removeCamBg.py --help.
usage: removeCamBg.py [-h] [-b BACKGROUND] [-u MASK_UPDATE] [-d MASK_DECAY]
[-m MASK_MAX] [-c CAMERA_ID] [--debug]
Read a webcam image and replace all background with pixels from a custom
image.
optional arguments:
-h, --help show this help message and exit
-b BACKGROUND, --background BACKGROUND
Filename of the background image. By default, it's a
noisy image
-u MASK_UPDATE, --mask-update MASK_UPDATE
Update the face mask every u frames
-d MASK_DECAY, --mask-decay MASK_DECAY
Upon losing the face, how fast shall the face fade
away. Between 0 and 1. 0 Means to delete mask
immediately, 1 means keep mask forever.
-m MASK_MAX, --mask-max MASK_MAX
Maximum blending factor of the detected face, between
0 and 1. 1 means the face is not transparent at all. 0
means face it not visible at all
-c CAMERA_ID, --camera-id CAMERA_ID
Id of the video capturing device to open. Will use 0
by default, that will open the default backend camera.
--debug If set, show the resulting image in a debug window
Holding a newspaper in front of the camera yields these results.
Input image:
Output image with some blending.
