OpenPose supports simultaneous detection of 135 keypoints across the body, face, and hands, uses a VGG-19 backbone network, and is difficult to run at real-time frame rates on devices without a GPU.