Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation | Read Paper on Bytez