This repo goes along with this Medium Blog Post. It contains code files for performing instance segmentation on live video capture devices.
It uses cv2 to start a video capture and renders the output which is either blurred or background substituted. More details about the implementation could be found in the medium blog post mentioned above.