Text this: Leveraging Temporal Down-Sampling Structure and Spatio-Temporal Fusion for Efficient Video Coding †.