Balancing Shared and Task-Specific Representations: A Hybrid Approach to Depth-Aware Video Panoptic Segmentation | IEEE Conference Publication | IEEE Xplore