Image stabilization using machine learning

This invention describes a method to smooth out shaky videos. It works by taking video frames and combining them with data from motion sensors, which track how the camera is moving (like tilting or rotating). A smart computer program, called a deep learning network, then uses this combined information to figure out how to adjust each video frame, making the final video appear much steadier. The claims specifically focus on using pitch, roll, and yaw motion data for video stabilization.