Directional Deep Embedding and Appearance Learning for Fast Video Object Segmentation | Read Paper on Bytez